diff --git "a/detection/mask_rcnn_beit_sbl_1568_1120_672_fpn_1x_coco_bs16.log" "b/detection/mask_rcnn_beit_sbl_1568_1120_672_fpn_1x_coco_bs16.log" new file mode 100644--- /dev/null +++ "b/detection/mask_rcnn_beit_sbl_1568_1120_672_fpn_1x_coco_bs16.log" @@ -0,0 +1,8948 @@ +2024-05-31 12:13:41,884 - mmdet - INFO - Environment info: +------------------------------------------------------------ +sys.platform: linux +Python: 3.9.19 (main, May 6 2024, 19:43:03) [GCC 11.2.0] +CUDA available: True +GPU 0,1,2,3,4,5,6,7: NVIDIA A100-SXM4-80GB +CUDA_HOME: /mnt/petrelfs/share/cuda-11.7/ +NVCC: Cuda compilation tools, release 11.7, V11.7.99 +GCC: gcc (GCC) 7.3.0 +PyTorch: 1.12.0+cu113 +PyTorch compiling details: PyTorch built with: + - GCC 9.3 + - C++ Version: 201402 + - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications + - Intel(R) MKL-DNN v2.6.0 (Git Hash 52b5f107dd9cf10910aaa19cb47f3abf9b349815) + - OpenMP 201511 (a.k.a. OpenMP 4.5) + - LAPACK is enabled (usually provided by MKL) + - NNPACK is enabled + - CPU capability usage: AVX2 + - CUDA Runtime 11.3 + - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75;-gencode;arch=compute_80,code=sm_80;-gencode;arch=compute_86,code=sm_86 + - CuDNN 8.3.2 (built against CUDA 11.5) + - Magma 2.5.2 + - 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= -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=1.12.0, 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: 0.13.0+cu113 +OpenCV: 4.9.0 +MMCV: 1.7.0 +MMCV Compiler: GCC 7.3 +MMCV CUDA Compiler: 11.7 +MMDetection: 2.25.3+c8d02d2 +------------------------------------------------------------ + +2024-05-31 12:13:43,408 - mmdet - INFO - Distributed training: True +2024-05-31 12:13:44,881 - mmdet - INFO - Config: +model = dict( + type='MaskRCNN', + backbone=dict( + type='PIIPThreeBranch', + n_points=4, + deform_num_heads=16, + cffn_ratio=0.25, + deform_ratio=0.5, + with_cffn=True, + interact_attn_type='deform', + interaction_drop_path_rate=0.4, + branch1=dict( + real_size=672, + pretrain_img_size=224, + patch_size=16, + pretrain_patch_size=16, + depth=24, + embed_dim=1024, + num_heads=16, + mlp_ratio=4, + qkv_bias=True, + drop_path_rate=0.4, + use_abs_pos_emb=False, + use_rel_pos_bias=True, + init_values=1e-06, + interaction_indexes=[[0, 1], [2, 3], [4, 5], [6, 7], [8, 9], + [10, 11], [12, 13], [14, 15], [16, 17], + [18, 19], [20, 21], [22, 23]], + pretrained='./pretrained/beitv2_large_patch16_224_pt1k_ft21k.pth', + window_attn=[ + True, True, True, True, True, True, True, True, True, True, + True, True, True, True, True, True, True, True, True, True, + True, True, True, True + ], + window_size=[ + 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, + 28, 28, 28, 28, 28, 28, 28, 28 + ], + img_norm_cfg=dict( + mean=[127.5, 127.5, 127.5], + std=[127.5, 127.5, 127.5], + to_rgb=True)), + branch2=dict( + real_size=1120, + pretrain_img_size=224, + patch_size=16, + pretrain_patch_size=16, + depth=12, + embed_dim=768, + num_heads=12, + mlp_ratio=4, + qkv_bias=True, + drop_path_rate=0.15, + use_abs_pos_emb=False, + use_rel_pos_bias=True, + init_values=1e-06, + interaction_indexes=[[0, 0], [1, 1], [2, 2], [3, 3], [4, 4], + [5, 5], [6, 6], [7, 7], [8, 8], [9, 9], + [10, 10], [11, 11]], + pretrained='./pretrained/beitv2_base_patch16_224_pt1k_ft21k.pth', + window_attn=[ + True, True, True, True, True, True, True, True, True, True, + True, True + ], + window_size=[28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28], + img_norm_cfg=dict( + mean=[127.5, 127.5, 127.5], + std=[127.5, 127.5, 127.5], + to_rgb=True)), + branch3=dict( + real_size=1568, + pretrain_img_size=224, + patch_size=16, + pretrain_patch_size=16, + depth=12, + embed_dim=384, + num_heads=6, + mlp_ratio=4, + qkv_bias=True, + drop_path_rate=0.05, + init_scale=1.0, + with_fpn=False, + interaction_indexes=[[0, 0], [1, 1], [2, 2], [3, 3], [4, 4], + [5, 5], [6, 6], [7, 7], [8, 8], [9, 9], + [10, 10], [11, 11]], + pretrained='./pretrained/deit_3_small_224_21k.pth', + window_attn=[ + True, True, True, True, True, True, True, True, True, True, + True, True + ], + window_size=[28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28], + use_flash_attn=True, + img_norm_cfg=dict( + mean=[127.5, 127.5, 127.5], + std=[127.5, 127.5, 127.5], + to_rgb=True))), + neck=dict( + type='FPN', + in_channels=[1024, 1024, 1024, 1024], + out_channels=256, + num_outs=5), + rpn_head=dict( + type='RPNHead', + in_channels=256, + feat_channels=256, + anchor_generator=dict( + type='AnchorGenerator', + scales=[8], + ratios=[0.5, 1.0, 2.0], + strides=[4, 8, 16, 32, 64]), + bbox_coder=dict( + type='DeltaXYWHBBoxCoder', + target_means=[0.0, 0.0, 0.0, 0.0], + target_stds=[1.0, 1.0, 1.0, 1.0]), + loss_cls=dict( + type='CrossEntropyLoss', use_sigmoid=True, loss_weight=1.0), + loss_bbox=dict(type='L1Loss', loss_weight=1.0)), + roi_head=dict( + type='StandardRoIHead', + bbox_roi_extractor=dict( + type='SingleRoIExtractor', + roi_layer=dict(type='RoIAlign', output_size=7, sampling_ratio=0), + out_channels=256, + featmap_strides=[4, 8, 16, 32]), + bbox_head=dict( + type='Shared2FCBBoxHead', + in_channels=256, + fc_out_channels=1024, + roi_feat_size=7, + num_classes=80, + bbox_coder=dict( + type='DeltaXYWHBBoxCoder', + target_means=[0.0, 0.0, 0.0, 0.0], + target_stds=[0.1, 0.1, 0.2, 0.2]), + reg_class_agnostic=False, + loss_cls=dict( + type='CrossEntropyLoss', use_sigmoid=False, loss_weight=1.0), + loss_bbox=dict(type='L1Loss', loss_weight=1.0)), + mask_roi_extractor=dict( + type='SingleRoIExtractor', + roi_layer=dict(type='RoIAlign', output_size=14, sampling_ratio=0), + out_channels=256, + featmap_strides=[4, 8, 16, 32]), + mask_head=dict( + type='FCNMaskHead', + num_convs=4, + in_channels=256, + conv_out_channels=256, + num_classes=80, + loss_mask=dict( + type='CrossEntropyLoss', use_mask=True, loss_weight=1.0))), + train_cfg=dict( + rpn=dict( + assigner=dict( + type='MaxIoUAssigner', + pos_iou_thr=0.7, + neg_iou_thr=0.3, + min_pos_iou=0.3, + match_low_quality=True, + ignore_iof_thr=-1), + sampler=dict( + type='RandomSampler', + num=256, + pos_fraction=0.5, + neg_pos_ub=-1, + add_gt_as_proposals=False), + allowed_border=-1, + pos_weight=-1, + debug=False), + rpn_proposal=dict( + nms_pre=2000, + max_per_img=1000, + nms=dict(type='nms', iou_threshold=0.7), + min_bbox_size=0), + rcnn=dict( + assigner=dict( + type='MaxIoUAssigner', + pos_iou_thr=0.5, + neg_iou_thr=0.5, + min_pos_iou=0.5, + match_low_quality=True, + ignore_iof_thr=-1), + sampler=dict( + type='RandomSampler', + num=512, + pos_fraction=0.25, + neg_pos_ub=-1, + add_gt_as_proposals=True), + mask_size=28, + pos_weight=-1, + debug=False)), + test_cfg=dict( + rpn=dict( + nms_pre=1000, + max_per_img=1000, + nms=dict(type='nms', iou_threshold=0.7), + min_bbox_size=0), + rcnn=dict( + score_thr=0.05, + nms=dict(type='nms', iou_threshold=0.5), + max_per_img=100, + mask_thr_binary=0.5))) +dataset_type = 'CocoDataset' +data_root = 'data/coco/' +img_norm_cfg = dict( + mean=[127.5, 127.5, 127.5], std=[127.5, 127.5, 127.5], to_rgb=True) +train_pipeline = [ + dict(type='LoadImageFromFile'), + dict(type='LoadAnnotations', with_bbox=True, with_mask=True), + dict(type='Resize', img_scale=(1568, 941), keep_ratio=True), + dict(type='RandomFlip', flip_ratio=0.5), + dict( + type='Normalize', + mean=[127.5, 127.5, 127.5], + std=[127.5, 127.5, 127.5], + to_rgb=True), + dict(type='Pad', size_divisor=224), + dict(type='DefaultFormatBundle'), + dict(type='Collect', keys=['img', 'gt_bboxes', 'gt_labels', 'gt_masks']) +] +test_pipeline = [ + dict(type='LoadImageFromFile'), + dict( + type='MultiScaleFlipAug', + img_scale=(1568, 941), + flip=False, + transforms=[ + dict(type='Resize', keep_ratio=True), + dict(type='RandomFlip'), + dict( + type='Normalize', + mean=[127.5, 127.5, 127.5], + std=[127.5, 127.5, 127.5], + to_rgb=True), + dict(type='Pad', size_divisor=224), + dict(type='ImageToTensor', keys=['img']), + dict(type='Collect', keys=['img']) + ]) +] +data = dict( + samples_per_gpu=2, + workers_per_gpu=2, + train=dict( + type='CocoDataset', + ann_file='data/coco/annotations/instances_train2017.json', + img_prefix='data/coco/train2017/', + pipeline=[ + dict(type='LoadImageFromFile'), + dict(type='LoadAnnotations', with_bbox=True, with_mask=True), + dict(type='Resize', img_scale=(1568, 941), keep_ratio=True), + dict(type='RandomFlip', flip_ratio=0.5), + dict( + type='Normalize', + mean=[127.5, 127.5, 127.5], + std=[127.5, 127.5, 127.5], + to_rgb=True), + dict(type='Pad', size_divisor=224), + dict(type='DefaultFormatBundle'), + dict( + type='Collect', + keys=['img', 'gt_bboxes', 'gt_labels', 'gt_masks']) + ]), + val=dict( + type='CocoDataset', + ann_file='data/coco/annotations/instances_val2017.json', + img_prefix='data/coco/val2017/', + pipeline=[ + dict(type='LoadImageFromFile'), + dict( + type='MultiScaleFlipAug', + img_scale=(1568, 941), + flip=False, + transforms=[ + dict(type='Resize', keep_ratio=True), + dict(type='RandomFlip'), + dict( + type='Normalize', + mean=[127.5, 127.5, 127.5], + std=[127.5, 127.5, 127.5], + to_rgb=True), + dict(type='Pad', size_divisor=224), + dict(type='ImageToTensor', keys=['img']), + dict(type='Collect', keys=['img']) + ]) + ]), + test=dict( + type='CocoDataset', + ann_file='data/coco/annotations/instances_val2017.json', + img_prefix='data/coco/val2017/', + pipeline=[ + dict(type='LoadImageFromFile'), + dict( + type='MultiScaleFlipAug', + img_scale=(1568, 941), + flip=False, + transforms=[ + dict(type='Resize', keep_ratio=True), + dict(type='RandomFlip'), + dict( + type='Normalize', + mean=[127.5, 127.5, 127.5], + std=[127.5, 127.5, 127.5], + to_rgb=True), + dict(type='Pad', size_divisor=224), + dict(type='ImageToTensor', keys=['img']), + dict(type='Collect', keys=['img']) + ]) + ])) +evaluation = dict(metric=['bbox', 'segm'], interval=1, save_best=None) +optimizer = dict( + type='AdamW', + lr=0.0001, + betas=(0.9, 0.999), + weight_decay=0.05, + constructor='CustomLayerDecayOptimizerConstructorMMDet', + paramwise_cfg=dict( + num_layers=24, layer_decay_rate=0.85, skip_stride=[2, 2])) +optimizer_config = dict(grad_clip=None) +lr_config = dict( + policy='step', + warmup='linear', + warmup_iters=500, + warmup_ratio=0.001, + step=[8, 11]) +runner = dict(type='EpochBasedRunner', max_epochs=12) +checkpoint_config = dict(interval=1, deepspeed=True, max_keep_ckpts=1) +log_config = dict(interval=50, hooks=[dict(type='TextLoggerHook')]) +custom_hooks = [dict(type='ToBFloat16HookMMDet', priority=49)] +dist_params = dict(backend='nccl') +log_level = 'INFO' +load_from = None +resume_from = None +workflow = [('train', 1)] +opencv_num_threads = 0 +mp_start_method = 'fork' +auto_scale_lr = dict(enable=False, base_batch_size=16) +deepspeed = True +deepspeed_config = 'zero_configs/adam_zero1_bf16.json' +custom_imports = dict( + imports=['mmdet.mmcv_custom'], allow_failed_imports=False) +work_dir = './work_dirs/mask_rcnn_beit_sbl_1568_1120_672_fpn_1x_coco_bs16' +auto_resume = True +gpu_ids = range(0, 8) + +2024-05-31 12:13:48,768 - mmdet - INFO - Set random seed to 1094034324, deterministic: False +2024-05-31 12:14:16,700 - mmdet - WARNING - The model and loaded state dict do not match exactly + +unexpected key in source state_dict: cls_token, fc_norm.weight, fc_norm.bias, head.weight, head.bias + +missing keys in source state_dict: blocks.0.attn.relative_position_index, blocks.1.attn.relative_position_index, blocks.2.attn.relative_position_index, blocks.3.attn.relative_position_index, blocks.4.attn.relative_position_index, blocks.5.attn.relative_position_index, blocks.6.attn.relative_position_index, blocks.7.attn.relative_position_index, blocks.8.attn.relative_position_index, blocks.9.attn.relative_position_index, blocks.10.attn.relative_position_index, blocks.11.attn.relative_position_index, blocks.12.attn.relative_position_index, blocks.13.attn.relative_position_index, blocks.14.attn.relative_position_index, blocks.15.attn.relative_position_index, blocks.16.attn.relative_position_index, blocks.17.attn.relative_position_index, blocks.18.attn.relative_position_index, blocks.19.attn.relative_position_index, blocks.20.attn.relative_position_index, blocks.21.attn.relative_position_index, blocks.22.attn.relative_position_index, blocks.23.attn.relative_position_index + +2024-05-31 12:14:31,160 - mmdet - WARNING - The model and loaded state dict do not match exactly + +unexpected key in source state_dict: cls_token, fc_norm.weight, fc_norm.bias, head.weight, head.bias + +missing keys in source state_dict: blocks.0.attn.relative_position_index, blocks.1.attn.relative_position_index, blocks.2.attn.relative_position_index, blocks.3.attn.relative_position_index, blocks.4.attn.relative_position_index, blocks.5.attn.relative_position_index, blocks.6.attn.relative_position_index, blocks.7.attn.relative_position_index, blocks.8.attn.relative_position_index, blocks.9.attn.relative_position_index, blocks.10.attn.relative_position_index, blocks.11.attn.relative_position_index + +2024-05-31 12:14:31,556 - mmdet - INFO - _IncompatibleKeys(missing_keys=[], unexpected_keys=['cls_token', 'norm.weight', 'norm.bias', 'head.weight', 'head.bias']) +2024-05-31 12:15:48,250 - mmdet - INFO - initialize FPN with init_cfg {'type': 'Xavier', 'layer': 'Conv2d', 'distribution': 'uniform'} +2024-05-31 12:15:48,737 - mmdet - INFO - initialize RPNHead with init_cfg {'type': 'Normal', 'layer': 'Conv2d', 'std': 0.01} +2024-05-31 12:15:48,797 - mmdet - INFO - initialize Shared2FCBBoxHead with init_cfg [{'type': 'Normal', 'std': 0.01, 'override': {'name': 'fc_cls'}}, {'type': 'Normal', 'std': 0.001, 'override': {'name': 'fc_reg'}}, {'type': 'Xavier', 'distribution': 'uniform', 'override': [{'name': 'shared_fcs'}, {'name': 'cls_fcs'}, {'name': 'reg_fcs'}]}] +Name of parameter - Initialization information + +backbone.w1 - torch.Size([]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.w2 - torch.Size([]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.w3 - torch.Size([]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.patch_embed.proj.weight - torch.Size([1024, 3, 16, 16]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.patch_embed.proj.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.0.gamma_1 - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.0.gamma_2 - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.0.norm1.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.0.norm1.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.0.attn.q_bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.0.attn.v_bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.0.attn.relative_position_bias_table - torch.Size([3025, 16]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.0.attn.qkv.weight - torch.Size([3072, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.0.attn.proj.weight - torch.Size([1024, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.0.attn.proj.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.0.norm2.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.0.norm2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.0.mlp.fc1.weight - torch.Size([4096, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.0.mlp.fc1.bias - torch.Size([4096]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.0.mlp.fc2.weight - torch.Size([1024, 4096]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.0.mlp.fc2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.1.gamma_1 - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.1.gamma_2 - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.1.norm1.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.1.norm1.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.1.attn.q_bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.1.attn.v_bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.1.attn.relative_position_bias_table - torch.Size([3025, 16]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.1.attn.qkv.weight - torch.Size([3072, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.1.attn.proj.weight - torch.Size([1024, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.1.attn.proj.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.1.norm2.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.1.norm2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.1.mlp.fc1.weight - torch.Size([4096, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.1.mlp.fc1.bias - torch.Size([4096]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.1.mlp.fc2.weight - torch.Size([1024, 4096]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.1.mlp.fc2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.2.gamma_1 - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.2.gamma_2 - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.2.norm1.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.2.norm1.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.2.attn.q_bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.2.attn.v_bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.2.attn.relative_position_bias_table - torch.Size([3025, 16]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.2.attn.qkv.weight - torch.Size([3072, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.2.attn.proj.weight - torch.Size([1024, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.2.attn.proj.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.2.norm2.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.2.norm2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.2.mlp.fc1.weight - torch.Size([4096, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.2.mlp.fc1.bias - torch.Size([4096]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.2.mlp.fc2.weight - torch.Size([1024, 4096]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.2.mlp.fc2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.3.gamma_1 - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.3.gamma_2 - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.3.norm1.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.3.norm1.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.3.attn.q_bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.3.attn.v_bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.3.attn.relative_position_bias_table - torch.Size([3025, 16]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.3.attn.qkv.weight - torch.Size([3072, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.3.attn.proj.weight - torch.Size([1024, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.3.attn.proj.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.3.norm2.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.3.norm2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.3.mlp.fc1.weight - torch.Size([4096, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.3.mlp.fc1.bias - torch.Size([4096]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.3.mlp.fc2.weight - torch.Size([1024, 4096]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.3.mlp.fc2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.4.gamma_1 - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.4.gamma_2 - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.4.norm1.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.4.norm1.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.4.attn.q_bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.4.attn.v_bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.4.attn.relative_position_bias_table - torch.Size([3025, 16]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.4.attn.qkv.weight - torch.Size([3072, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.4.attn.proj.weight - torch.Size([1024, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.4.attn.proj.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.4.norm2.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.4.norm2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.4.mlp.fc1.weight - torch.Size([4096, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.4.mlp.fc1.bias - torch.Size([4096]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.4.mlp.fc2.weight - torch.Size([1024, 4096]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.4.mlp.fc2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.5.gamma_1 - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.5.gamma_2 - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.5.norm1.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.5.norm1.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.5.attn.q_bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.5.attn.v_bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.5.attn.relative_position_bias_table - torch.Size([3025, 16]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.5.attn.qkv.weight - torch.Size([3072, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.5.attn.proj.weight - torch.Size([1024, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.5.attn.proj.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.5.norm2.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.5.norm2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.5.mlp.fc1.weight - torch.Size([4096, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.5.mlp.fc1.bias - torch.Size([4096]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.5.mlp.fc2.weight - torch.Size([1024, 4096]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.5.mlp.fc2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.6.gamma_1 - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.6.gamma_2 - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.6.norm1.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.6.norm1.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.6.attn.q_bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.6.attn.v_bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.6.attn.relative_position_bias_table - torch.Size([3025, 16]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.6.attn.qkv.weight - torch.Size([3072, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.6.attn.proj.weight - torch.Size([1024, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.6.attn.proj.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.6.norm2.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.6.norm2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.6.mlp.fc1.weight - torch.Size([4096, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.6.mlp.fc1.bias - torch.Size([4096]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.6.mlp.fc2.weight - torch.Size([1024, 4096]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.6.mlp.fc2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.7.gamma_1 - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.7.gamma_2 - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.7.norm1.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.7.norm1.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.7.attn.q_bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.7.attn.v_bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.7.attn.relative_position_bias_table - torch.Size([3025, 16]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.7.attn.qkv.weight - torch.Size([3072, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.7.attn.proj.weight - torch.Size([1024, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.7.attn.proj.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.7.norm2.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.7.norm2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.7.mlp.fc1.weight - torch.Size([4096, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.7.mlp.fc1.bias - torch.Size([4096]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.7.mlp.fc2.weight - torch.Size([1024, 4096]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.7.mlp.fc2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.8.gamma_1 - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.8.gamma_2 - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.8.norm1.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.8.norm1.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.8.attn.q_bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.8.attn.v_bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.8.attn.relative_position_bias_table - torch.Size([3025, 16]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.8.attn.qkv.weight - torch.Size([3072, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.8.attn.proj.weight - torch.Size([1024, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.8.attn.proj.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.8.norm2.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.8.norm2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.8.mlp.fc1.weight - torch.Size([4096, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.8.mlp.fc1.bias - torch.Size([4096]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.8.mlp.fc2.weight - torch.Size([1024, 4096]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.8.mlp.fc2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.9.gamma_1 - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.9.gamma_2 - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.9.norm1.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.9.norm1.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.9.attn.q_bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.9.attn.v_bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.9.attn.relative_position_bias_table - torch.Size([3025, 16]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.9.attn.qkv.weight - torch.Size([3072, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.9.attn.proj.weight - torch.Size([1024, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.9.attn.proj.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.9.norm2.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.9.norm2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.9.mlp.fc1.weight - torch.Size([4096, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.9.mlp.fc1.bias - torch.Size([4096]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.9.mlp.fc2.weight - torch.Size([1024, 4096]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.9.mlp.fc2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.10.gamma_1 - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.10.gamma_2 - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.10.norm1.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.10.norm1.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.10.attn.q_bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.10.attn.v_bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.10.attn.relative_position_bias_table - torch.Size([3025, 16]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.10.attn.qkv.weight - torch.Size([3072, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.10.attn.proj.weight - torch.Size([1024, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.10.attn.proj.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.10.norm2.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.10.norm2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.10.mlp.fc1.weight - torch.Size([4096, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.10.mlp.fc1.bias - torch.Size([4096]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.10.mlp.fc2.weight - torch.Size([1024, 4096]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.10.mlp.fc2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.11.gamma_1 - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.11.gamma_2 - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.11.norm1.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.11.norm1.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.11.attn.q_bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.11.attn.v_bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.11.attn.relative_position_bias_table - torch.Size([3025, 16]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.11.attn.qkv.weight - torch.Size([3072, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.11.attn.proj.weight - torch.Size([1024, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.11.attn.proj.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.11.norm2.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.11.norm2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.11.mlp.fc1.weight - torch.Size([4096, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.11.mlp.fc1.bias - torch.Size([4096]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.11.mlp.fc2.weight - torch.Size([1024, 4096]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.11.mlp.fc2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.12.gamma_1 - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.12.gamma_2 - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.12.norm1.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.12.norm1.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.12.attn.q_bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.12.attn.v_bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.12.attn.relative_position_bias_table - torch.Size([3025, 16]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.12.attn.qkv.weight - torch.Size([3072, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.12.attn.proj.weight - torch.Size([1024, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.12.attn.proj.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.12.norm2.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.12.norm2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.12.mlp.fc1.weight - torch.Size([4096, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.12.mlp.fc1.bias - torch.Size([4096]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.12.mlp.fc2.weight - torch.Size([1024, 4096]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.12.mlp.fc2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.13.gamma_1 - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.13.gamma_2 - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.13.norm1.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.13.norm1.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.13.attn.q_bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.13.attn.v_bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.13.attn.relative_position_bias_table - torch.Size([3025, 16]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.13.attn.qkv.weight - torch.Size([3072, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.13.attn.proj.weight - torch.Size([1024, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.13.attn.proj.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.13.norm2.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.13.norm2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.13.mlp.fc1.weight - torch.Size([4096, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.13.mlp.fc1.bias - torch.Size([4096]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.13.mlp.fc2.weight - torch.Size([1024, 4096]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.13.mlp.fc2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.14.gamma_1 - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.14.gamma_2 - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.14.norm1.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.14.norm1.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.14.attn.q_bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.14.attn.v_bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.14.attn.relative_position_bias_table - torch.Size([3025, 16]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.14.attn.qkv.weight - torch.Size([3072, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.14.attn.proj.weight - torch.Size([1024, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.14.attn.proj.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.14.norm2.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.14.norm2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.14.mlp.fc1.weight - torch.Size([4096, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.14.mlp.fc1.bias - torch.Size([4096]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.14.mlp.fc2.weight - torch.Size([1024, 4096]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.14.mlp.fc2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.15.gamma_1 - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.15.gamma_2 - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.15.norm1.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.15.norm1.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.15.attn.q_bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.15.attn.v_bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.15.attn.relative_position_bias_table - torch.Size([3025, 16]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.15.attn.qkv.weight - torch.Size([3072, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.15.attn.proj.weight - torch.Size([1024, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.15.attn.proj.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.15.norm2.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.15.norm2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.15.mlp.fc1.weight - torch.Size([4096, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.15.mlp.fc1.bias - torch.Size([4096]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.15.mlp.fc2.weight - torch.Size([1024, 4096]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.15.mlp.fc2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.16.gamma_1 - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.16.gamma_2 - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.16.norm1.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.16.norm1.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.16.attn.q_bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.16.attn.v_bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.16.attn.relative_position_bias_table - torch.Size([3025, 16]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.16.attn.qkv.weight - torch.Size([3072, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.16.attn.proj.weight - torch.Size([1024, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.16.attn.proj.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.16.norm2.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.16.norm2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.16.mlp.fc1.weight - torch.Size([4096, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.16.mlp.fc1.bias - torch.Size([4096]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.16.mlp.fc2.weight - torch.Size([1024, 4096]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.16.mlp.fc2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.17.gamma_1 - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.17.gamma_2 - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.17.norm1.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.17.norm1.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.17.attn.q_bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.17.attn.v_bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.17.attn.relative_position_bias_table - torch.Size([3025, 16]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.17.attn.qkv.weight - torch.Size([3072, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.17.attn.proj.weight - torch.Size([1024, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.17.attn.proj.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.17.norm2.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.17.norm2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.17.mlp.fc1.weight - torch.Size([4096, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.17.mlp.fc1.bias - torch.Size([4096]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.17.mlp.fc2.weight - torch.Size([1024, 4096]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.17.mlp.fc2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.18.gamma_1 - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.18.gamma_2 - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.18.norm1.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.18.norm1.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.18.attn.q_bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.18.attn.v_bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.18.attn.relative_position_bias_table - torch.Size([3025, 16]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.18.attn.qkv.weight - torch.Size([3072, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.18.attn.proj.weight - torch.Size([1024, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.18.attn.proj.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.18.norm2.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.18.norm2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.18.mlp.fc1.weight - torch.Size([4096, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.18.mlp.fc1.bias - torch.Size([4096]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.18.mlp.fc2.weight - torch.Size([1024, 4096]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.18.mlp.fc2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.19.gamma_1 - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.19.gamma_2 - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.19.norm1.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.19.norm1.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.19.attn.q_bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.19.attn.v_bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.19.attn.relative_position_bias_table - torch.Size([3025, 16]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.19.attn.qkv.weight - torch.Size([3072, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.19.attn.proj.weight - torch.Size([1024, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.19.attn.proj.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.19.norm2.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.19.norm2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.19.mlp.fc1.weight - torch.Size([4096, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.19.mlp.fc1.bias - torch.Size([4096]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.19.mlp.fc2.weight - torch.Size([1024, 4096]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.19.mlp.fc2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.20.gamma_1 - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.20.gamma_2 - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.20.norm1.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.20.norm1.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.20.attn.q_bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.20.attn.v_bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.20.attn.relative_position_bias_table - torch.Size([3025, 16]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.20.attn.qkv.weight - torch.Size([3072, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.20.attn.proj.weight - torch.Size([1024, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.20.attn.proj.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.20.norm2.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.20.norm2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.20.mlp.fc1.weight - torch.Size([4096, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.20.mlp.fc1.bias - torch.Size([4096]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.20.mlp.fc2.weight - torch.Size([1024, 4096]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.20.mlp.fc2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.21.gamma_1 - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.21.gamma_2 - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.21.norm1.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.21.norm1.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.21.attn.q_bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.21.attn.v_bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.21.attn.relative_position_bias_table - torch.Size([3025, 16]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.21.attn.qkv.weight - torch.Size([3072, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.21.attn.proj.weight - torch.Size([1024, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.21.attn.proj.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.21.norm2.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.21.norm2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.21.mlp.fc1.weight - torch.Size([4096, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.21.mlp.fc1.bias - torch.Size([4096]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.21.mlp.fc2.weight - torch.Size([1024, 4096]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.21.mlp.fc2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.22.gamma_1 - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.22.gamma_2 - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.22.norm1.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.22.norm1.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.22.attn.q_bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.22.attn.v_bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.22.attn.relative_position_bias_table - torch.Size([3025, 16]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.22.attn.qkv.weight - torch.Size([3072, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.22.attn.proj.weight - torch.Size([1024, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.22.attn.proj.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.22.norm2.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.22.norm2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.22.mlp.fc1.weight - torch.Size([4096, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.22.mlp.fc1.bias - torch.Size([4096]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.22.mlp.fc2.weight - torch.Size([1024, 4096]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.22.mlp.fc2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.23.gamma_1 - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.23.gamma_2 - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.23.norm1.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.23.norm1.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.23.attn.q_bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.23.attn.v_bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.23.attn.relative_position_bias_table - torch.Size([3025, 16]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.23.attn.qkv.weight - torch.Size([3072, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.23.attn.proj.weight - torch.Size([1024, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.23.attn.proj.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.23.norm2.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.23.norm2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.23.mlp.fc1.weight - torch.Size([4096, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.23.mlp.fc1.bias - torch.Size([4096]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.23.mlp.fc2.weight - torch.Size([1024, 4096]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch1.blocks.23.mlp.fc2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.patch_embed.proj.weight - torch.Size([768, 3, 16, 16]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.patch_embed.proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.0.gamma_1 - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.0.gamma_2 - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.0.norm1.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.0.norm1.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.0.attn.q_bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.0.attn.v_bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.0.attn.relative_position_bias_table - torch.Size([3025, 12]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.0.attn.qkv.weight - torch.Size([2304, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.0.attn.proj.weight - torch.Size([768, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.0.attn.proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.0.norm2.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.0.norm2.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.0.mlp.fc1.weight - torch.Size([3072, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.0.mlp.fc1.bias - torch.Size([3072]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.0.mlp.fc2.weight - torch.Size([768, 3072]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.0.mlp.fc2.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.1.gamma_1 - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.1.gamma_2 - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.1.norm1.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.1.norm1.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.1.attn.q_bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.1.attn.v_bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.1.attn.relative_position_bias_table - torch.Size([3025, 12]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.1.attn.qkv.weight - torch.Size([2304, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.1.attn.proj.weight - torch.Size([768, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.1.attn.proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.1.norm2.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.1.norm2.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.1.mlp.fc1.weight - torch.Size([3072, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.1.mlp.fc1.bias - torch.Size([3072]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.1.mlp.fc2.weight - torch.Size([768, 3072]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.1.mlp.fc2.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.2.gamma_1 - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.2.gamma_2 - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.2.norm1.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.2.norm1.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.2.attn.q_bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.2.attn.v_bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.2.attn.relative_position_bias_table - torch.Size([3025, 12]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.2.attn.qkv.weight - torch.Size([2304, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.2.attn.proj.weight - torch.Size([768, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.2.attn.proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.2.norm2.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.2.norm2.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.2.mlp.fc1.weight - torch.Size([3072, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.2.mlp.fc1.bias - torch.Size([3072]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.2.mlp.fc2.weight - torch.Size([768, 3072]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.2.mlp.fc2.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.3.gamma_1 - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.3.gamma_2 - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.3.norm1.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.3.norm1.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.3.attn.q_bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.3.attn.v_bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.3.attn.relative_position_bias_table - torch.Size([3025, 12]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.3.attn.qkv.weight - torch.Size([2304, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.3.attn.proj.weight - torch.Size([768, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.3.attn.proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.3.norm2.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.3.norm2.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.3.mlp.fc1.weight - torch.Size([3072, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.3.mlp.fc1.bias - torch.Size([3072]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.3.mlp.fc2.weight - torch.Size([768, 3072]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.3.mlp.fc2.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.4.gamma_1 - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.4.gamma_2 - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.4.norm1.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.4.norm1.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.4.attn.q_bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.4.attn.v_bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.4.attn.relative_position_bias_table - torch.Size([3025, 12]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.4.attn.qkv.weight - torch.Size([2304, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.4.attn.proj.weight - torch.Size([768, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.4.attn.proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.4.norm2.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.4.norm2.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.4.mlp.fc1.weight - torch.Size([3072, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.4.mlp.fc1.bias - torch.Size([3072]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.4.mlp.fc2.weight - torch.Size([768, 3072]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.4.mlp.fc2.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.5.gamma_1 - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.5.gamma_2 - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.5.norm1.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.5.norm1.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.5.attn.q_bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.5.attn.v_bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.5.attn.relative_position_bias_table - torch.Size([3025, 12]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.5.attn.qkv.weight - torch.Size([2304, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.5.attn.proj.weight - torch.Size([768, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.5.attn.proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.5.norm2.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.5.norm2.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.5.mlp.fc1.weight - torch.Size([3072, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.5.mlp.fc1.bias - torch.Size([3072]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.5.mlp.fc2.weight - torch.Size([768, 3072]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.5.mlp.fc2.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.6.gamma_1 - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.6.gamma_2 - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.6.norm1.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.6.norm1.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.6.attn.q_bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.6.attn.v_bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.6.attn.relative_position_bias_table - torch.Size([3025, 12]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.6.attn.qkv.weight - torch.Size([2304, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.6.attn.proj.weight - torch.Size([768, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.6.attn.proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.6.norm2.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.6.norm2.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.6.mlp.fc1.weight - torch.Size([3072, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.6.mlp.fc1.bias - torch.Size([3072]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.6.mlp.fc2.weight - torch.Size([768, 3072]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.6.mlp.fc2.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.7.gamma_1 - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.7.gamma_2 - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.7.norm1.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.7.norm1.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.7.attn.q_bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.7.attn.v_bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.7.attn.relative_position_bias_table - torch.Size([3025, 12]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.7.attn.qkv.weight - torch.Size([2304, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.7.attn.proj.weight - torch.Size([768, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.7.attn.proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.7.norm2.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.7.norm2.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.7.mlp.fc1.weight - torch.Size([3072, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.7.mlp.fc1.bias - torch.Size([3072]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.7.mlp.fc2.weight - torch.Size([768, 3072]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.7.mlp.fc2.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.8.gamma_1 - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.8.gamma_2 - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.8.norm1.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.8.norm1.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.8.attn.q_bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.8.attn.v_bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.8.attn.relative_position_bias_table - torch.Size([3025, 12]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.8.attn.qkv.weight - torch.Size([2304, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.8.attn.proj.weight - torch.Size([768, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.8.attn.proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.8.norm2.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.8.norm2.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.8.mlp.fc1.weight - torch.Size([3072, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.8.mlp.fc1.bias - torch.Size([3072]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.8.mlp.fc2.weight - torch.Size([768, 3072]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.8.mlp.fc2.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.9.gamma_1 - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.9.gamma_2 - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.9.norm1.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.9.norm1.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.9.attn.q_bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.9.attn.v_bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.9.attn.relative_position_bias_table - torch.Size([3025, 12]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.9.attn.qkv.weight - torch.Size([2304, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.9.attn.proj.weight - torch.Size([768, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.9.attn.proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.9.norm2.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.9.norm2.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.9.mlp.fc1.weight - torch.Size([3072, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.9.mlp.fc1.bias - torch.Size([3072]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.9.mlp.fc2.weight - torch.Size([768, 3072]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.9.mlp.fc2.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.10.gamma_1 - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.10.gamma_2 - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.10.norm1.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.10.norm1.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.10.attn.q_bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.10.attn.v_bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.10.attn.relative_position_bias_table - torch.Size([3025, 12]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.10.attn.qkv.weight - torch.Size([2304, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.10.attn.proj.weight - torch.Size([768, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.10.attn.proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.10.norm2.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.10.norm2.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.10.mlp.fc1.weight - torch.Size([3072, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.10.mlp.fc1.bias - torch.Size([3072]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.10.mlp.fc2.weight - torch.Size([768, 3072]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.10.mlp.fc2.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.11.gamma_1 - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.11.gamma_2 - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.11.norm1.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.11.norm1.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.11.attn.q_bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.11.attn.v_bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.11.attn.relative_position_bias_table - torch.Size([3025, 12]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.11.attn.qkv.weight - torch.Size([2304, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.11.attn.proj.weight - torch.Size([768, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.11.attn.proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.11.norm2.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.11.norm2.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.11.mlp.fc1.weight - torch.Size([3072, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.11.mlp.fc1.bias - torch.Size([3072]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.11.mlp.fc2.weight - torch.Size([768, 3072]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch2.blocks.11.mlp.fc2.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.pos_embed - torch.Size([1, 196, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.patch_embed.proj.weight - torch.Size([384, 3, 16, 16]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.patch_embed.proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.0.gamma_1 - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.0.gamma_2 - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.0.norm1.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.0.norm1.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.0.attn.qkv.weight - torch.Size([1152, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.0.attn.qkv.bias - torch.Size([1152]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.0.attn.proj.weight - torch.Size([384, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.0.attn.proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.0.norm2.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.0.norm2.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.0.mlp.fc1.weight - torch.Size([1536, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.0.mlp.fc1.bias - torch.Size([1536]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.0.mlp.fc2.weight - torch.Size([384, 1536]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.0.mlp.fc2.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.1.gamma_1 - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.1.gamma_2 - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.1.norm1.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.1.norm1.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.1.attn.qkv.weight - torch.Size([1152, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.1.attn.qkv.bias - torch.Size([1152]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.1.attn.proj.weight - torch.Size([384, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.1.attn.proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.1.norm2.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.1.norm2.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.1.mlp.fc1.weight - torch.Size([1536, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.1.mlp.fc1.bias - torch.Size([1536]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.1.mlp.fc2.weight - torch.Size([384, 1536]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.1.mlp.fc2.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.2.gamma_1 - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.2.gamma_2 - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.2.norm1.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.2.norm1.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.2.attn.qkv.weight - torch.Size([1152, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.2.attn.qkv.bias - torch.Size([1152]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.2.attn.proj.weight - torch.Size([384, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.2.attn.proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.2.norm2.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.2.norm2.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.2.mlp.fc1.weight - torch.Size([1536, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.2.mlp.fc1.bias - torch.Size([1536]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.2.mlp.fc2.weight - torch.Size([384, 1536]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.2.mlp.fc2.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.3.gamma_1 - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.3.gamma_2 - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.3.norm1.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.3.norm1.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.3.attn.qkv.weight - torch.Size([1152, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.3.attn.qkv.bias - torch.Size([1152]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.3.attn.proj.weight - torch.Size([384, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.3.attn.proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.3.norm2.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.3.norm2.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.3.mlp.fc1.weight - torch.Size([1536, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.3.mlp.fc1.bias - torch.Size([1536]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.3.mlp.fc2.weight - torch.Size([384, 1536]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.3.mlp.fc2.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.4.gamma_1 - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.4.gamma_2 - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.4.norm1.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.4.norm1.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.4.attn.qkv.weight - torch.Size([1152, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.4.attn.qkv.bias - torch.Size([1152]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.4.attn.proj.weight - torch.Size([384, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.4.attn.proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.4.norm2.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.4.norm2.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.4.mlp.fc1.weight - torch.Size([1536, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.4.mlp.fc1.bias - torch.Size([1536]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.4.mlp.fc2.weight - torch.Size([384, 1536]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.4.mlp.fc2.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.5.gamma_1 - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.5.gamma_2 - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.5.norm1.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.5.norm1.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.5.attn.qkv.weight - torch.Size([1152, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.5.attn.qkv.bias - torch.Size([1152]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.5.attn.proj.weight - torch.Size([384, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.5.attn.proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.5.norm2.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.5.norm2.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.5.mlp.fc1.weight - torch.Size([1536, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.5.mlp.fc1.bias - torch.Size([1536]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.5.mlp.fc2.weight - torch.Size([384, 1536]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.5.mlp.fc2.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.6.gamma_1 - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.6.gamma_2 - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.6.norm1.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.6.norm1.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.6.attn.qkv.weight - torch.Size([1152, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.6.attn.qkv.bias - torch.Size([1152]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.6.attn.proj.weight - torch.Size([384, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.6.attn.proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.6.norm2.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.6.norm2.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.6.mlp.fc1.weight - torch.Size([1536, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.6.mlp.fc1.bias - torch.Size([1536]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.6.mlp.fc2.weight - torch.Size([384, 1536]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.6.mlp.fc2.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.7.gamma_1 - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.7.gamma_2 - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.7.norm1.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.7.norm1.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.7.attn.qkv.weight - torch.Size([1152, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.7.attn.qkv.bias - torch.Size([1152]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.7.attn.proj.weight - torch.Size([384, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.7.attn.proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.7.norm2.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.7.norm2.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.7.mlp.fc1.weight - torch.Size([1536, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.7.mlp.fc1.bias - torch.Size([1536]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.7.mlp.fc2.weight - torch.Size([384, 1536]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.7.mlp.fc2.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.8.gamma_1 - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.8.gamma_2 - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.8.norm1.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.8.norm1.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.8.attn.qkv.weight - torch.Size([1152, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.8.attn.qkv.bias - torch.Size([1152]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.8.attn.proj.weight - torch.Size([384, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.8.attn.proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.8.norm2.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.8.norm2.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.8.mlp.fc1.weight - torch.Size([1536, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.8.mlp.fc1.bias - torch.Size([1536]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.8.mlp.fc2.weight - torch.Size([384, 1536]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.8.mlp.fc2.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.9.gamma_1 - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.9.gamma_2 - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.9.norm1.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.9.norm1.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.9.attn.qkv.weight - torch.Size([1152, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.9.attn.qkv.bias - torch.Size([1152]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.9.attn.proj.weight - torch.Size([384, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.9.attn.proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.9.norm2.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.9.norm2.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.9.mlp.fc1.weight - torch.Size([1536, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.9.mlp.fc1.bias - torch.Size([1536]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.9.mlp.fc2.weight - torch.Size([384, 1536]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.9.mlp.fc2.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.10.gamma_1 - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.10.gamma_2 - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.10.norm1.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.10.norm1.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.10.attn.qkv.weight - torch.Size([1152, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.10.attn.qkv.bias - torch.Size([1152]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.10.attn.proj.weight - torch.Size([384, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.10.attn.proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.10.norm2.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.10.norm2.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.10.mlp.fc1.weight - torch.Size([1536, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.10.mlp.fc1.bias - torch.Size([1536]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.10.mlp.fc2.weight - torch.Size([384, 1536]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.10.mlp.fc2.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.11.gamma_1 - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.11.gamma_2 - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.11.norm1.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.11.norm1.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.11.attn.qkv.weight - torch.Size([1152, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.11.attn.qkv.bias - torch.Size([1152]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.11.attn.proj.weight - torch.Size([384, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.11.attn.proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.11.norm2.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.11.norm2.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.11.mlp.fc1.weight - torch.Size([1536, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.11.mlp.fc1.bias - torch.Size([1536]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.11.mlp.fc2.weight - torch.Size([384, 1536]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.branch3.blocks.11.mlp.fc2.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_12.branch2to1_proj.weight - torch.Size([1024, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_12.branch2to1_proj.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_12.branch2to1_injector.ca_gamma - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_12.branch2to1_injector.cffn_gamma - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_12.branch2to1_injector.query_norm.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_12.branch2to1_injector.query_norm.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_12.branch2to1_injector.feat_norm.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_12.branch2to1_injector.feat_norm.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_12.branch2to1_injector.attn.sampling_offsets.weight - torch.Size([128, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_12.branch2to1_injector.attn.sampling_offsets.bias - torch.Size([128]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_12.branch2to1_injector.attn.attention_weights.weight - torch.Size([64, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_12.branch2to1_injector.attn.attention_weights.bias - torch.Size([64]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_12.branch2to1_injector.attn.value_proj.weight - torch.Size([512, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_12.branch2to1_injector.attn.value_proj.bias - torch.Size([512]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_12.branch2to1_injector.attn.output_proj.weight - torch.Size([1024, 512]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_12.branch2to1_injector.attn.output_proj.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_12.branch2to1_injector.ffn.fc1.weight - torch.Size([256, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_12.branch2to1_injector.ffn.fc1.bias - torch.Size([256]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_12.branch2to1_injector.ffn.dwconv.dwconv.weight - torch.Size([256, 1, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_12.branch2to1_injector.ffn.dwconv.dwconv.bias - torch.Size([256]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_12.branch2to1_injector.ffn.fc2.weight - torch.Size([1024, 256]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_12.branch2to1_injector.ffn.fc2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_12.branch2to1_injector.ffn_norm.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_12.branch2to1_injector.ffn_norm.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_12.branch1to2_proj.weight - torch.Size([768, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_12.branch1to2_proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_12.branch1to2_injector.ca_gamma - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_12.branch1to2_injector.cffn_gamma - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_12.branch1to2_injector.query_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_12.branch1to2_injector.query_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_12.branch1to2_injector.feat_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_12.branch1to2_injector.feat_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_12.branch1to2_injector.attn.sampling_offsets.weight - torch.Size([128, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_12.branch1to2_injector.attn.sampling_offsets.bias - torch.Size([128]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_12.branch1to2_injector.attn.attention_weights.weight - torch.Size([64, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_12.branch1to2_injector.attn.attention_weights.bias - torch.Size([64]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_12.branch1to2_injector.attn.value_proj.weight - torch.Size([384, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_12.branch1to2_injector.attn.value_proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_12.branch1to2_injector.attn.output_proj.weight - torch.Size([768, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_12.branch1to2_injector.attn.output_proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_12.branch1to2_injector.ffn.fc1.weight - torch.Size([192, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_12.branch1to2_injector.ffn.fc1.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_12.branch1to2_injector.ffn.dwconv.dwconv.weight - torch.Size([192, 1, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_12.branch1to2_injector.ffn.dwconv.dwconv.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_12.branch1to2_injector.ffn.fc2.weight - torch.Size([768, 192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_12.branch1to2_injector.ffn.fc2.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_12.branch1to2_injector.ffn_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_12.branch1to2_injector.ffn_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_23.branch2to1_proj.weight - torch.Size([768, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_23.branch2to1_proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_23.branch2to1_injector.ca_gamma - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_23.branch2to1_injector.cffn_gamma - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_23.branch2to1_injector.query_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_23.branch2to1_injector.query_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_23.branch2to1_injector.feat_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_23.branch2to1_injector.feat_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_23.branch2to1_injector.attn.sampling_offsets.weight - torch.Size([128, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_23.branch2to1_injector.attn.sampling_offsets.bias - torch.Size([128]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_23.branch2to1_injector.attn.attention_weights.weight - torch.Size([64, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_23.branch2to1_injector.attn.attention_weights.bias - torch.Size([64]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_23.branch2to1_injector.attn.value_proj.weight - torch.Size([384, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_23.branch2to1_injector.attn.value_proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_23.branch2to1_injector.attn.output_proj.weight - torch.Size([768, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_23.branch2to1_injector.attn.output_proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_23.branch2to1_injector.ffn.fc1.weight - torch.Size([192, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_23.branch2to1_injector.ffn.fc1.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_23.branch2to1_injector.ffn.dwconv.dwconv.weight - torch.Size([192, 1, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_23.branch2to1_injector.ffn.dwconv.dwconv.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_23.branch2to1_injector.ffn.fc2.weight - torch.Size([768, 192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_23.branch2to1_injector.ffn.fc2.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_23.branch2to1_injector.ffn_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_23.branch2to1_injector.ffn_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_23.branch1to2_proj.weight - torch.Size([384, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_23.branch1to2_proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_23.branch1to2_injector.ca_gamma - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_23.branch1to2_injector.cffn_gamma - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_23.branch1to2_injector.query_norm.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_23.branch1to2_injector.query_norm.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_23.branch1to2_injector.feat_norm.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_23.branch1to2_injector.feat_norm.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_23.branch1to2_injector.attn.sampling_offsets.weight - torch.Size([128, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_23.branch1to2_injector.attn.sampling_offsets.bias - torch.Size([128]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_23.branch1to2_injector.attn.attention_weights.weight - torch.Size([64, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_23.branch1to2_injector.attn.attention_weights.bias - torch.Size([64]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_23.branch1to2_injector.attn.value_proj.weight - torch.Size([192, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_23.branch1to2_injector.attn.value_proj.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_23.branch1to2_injector.attn.output_proj.weight - torch.Size([384, 192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_23.branch1to2_injector.attn.output_proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_23.branch1to2_injector.ffn.fc1.weight - torch.Size([96, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_23.branch1to2_injector.ffn.fc1.bias - torch.Size([96]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_23.branch1to2_injector.ffn.dwconv.dwconv.weight - torch.Size([96, 1, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_23.branch1to2_injector.ffn.dwconv.dwconv.bias - torch.Size([96]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_23.branch1to2_injector.ffn.fc2.weight - torch.Size([384, 96]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_23.branch1to2_injector.ffn.fc2.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_23.branch1to2_injector.ffn_norm.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.0.interaction_units_23.branch1to2_injector.ffn_norm.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_12.branch2to1_proj.weight - torch.Size([1024, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_12.branch2to1_proj.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_12.branch2to1_injector.ca_gamma - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_12.branch2to1_injector.cffn_gamma - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_12.branch2to1_injector.query_norm.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_12.branch2to1_injector.query_norm.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_12.branch2to1_injector.feat_norm.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_12.branch2to1_injector.feat_norm.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_12.branch2to1_injector.attn.sampling_offsets.weight - torch.Size([128, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_12.branch2to1_injector.attn.sampling_offsets.bias - torch.Size([128]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_12.branch2to1_injector.attn.attention_weights.weight - torch.Size([64, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_12.branch2to1_injector.attn.attention_weights.bias - torch.Size([64]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_12.branch2to1_injector.attn.value_proj.weight - torch.Size([512, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_12.branch2to1_injector.attn.value_proj.bias - torch.Size([512]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_12.branch2to1_injector.attn.output_proj.weight - torch.Size([1024, 512]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_12.branch2to1_injector.attn.output_proj.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_12.branch2to1_injector.ffn.fc1.weight - torch.Size([256, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_12.branch2to1_injector.ffn.fc1.bias - torch.Size([256]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_12.branch2to1_injector.ffn.dwconv.dwconv.weight - torch.Size([256, 1, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_12.branch2to1_injector.ffn.dwconv.dwconv.bias - torch.Size([256]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_12.branch2to1_injector.ffn.fc2.weight - torch.Size([1024, 256]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_12.branch2to1_injector.ffn.fc2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_12.branch2to1_injector.ffn_norm.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_12.branch2to1_injector.ffn_norm.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_12.branch1to2_proj.weight - torch.Size([768, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_12.branch1to2_proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_12.branch1to2_injector.ca_gamma - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_12.branch1to2_injector.cffn_gamma - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_12.branch1to2_injector.query_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_12.branch1to2_injector.query_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_12.branch1to2_injector.feat_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_12.branch1to2_injector.feat_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_12.branch1to2_injector.attn.sampling_offsets.weight - torch.Size([128, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_12.branch1to2_injector.attn.sampling_offsets.bias - torch.Size([128]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_12.branch1to2_injector.attn.attention_weights.weight - torch.Size([64, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_12.branch1to2_injector.attn.attention_weights.bias - torch.Size([64]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_12.branch1to2_injector.attn.value_proj.weight - torch.Size([384, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_12.branch1to2_injector.attn.value_proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_12.branch1to2_injector.attn.output_proj.weight - torch.Size([768, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_12.branch1to2_injector.attn.output_proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_12.branch1to2_injector.ffn.fc1.weight - torch.Size([192, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_12.branch1to2_injector.ffn.fc1.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_12.branch1to2_injector.ffn.dwconv.dwconv.weight - torch.Size([192, 1, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_12.branch1to2_injector.ffn.dwconv.dwconv.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_12.branch1to2_injector.ffn.fc2.weight - torch.Size([768, 192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_12.branch1to2_injector.ffn.fc2.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_12.branch1to2_injector.ffn_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_12.branch1to2_injector.ffn_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_23.branch2to1_proj.weight - torch.Size([768, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_23.branch2to1_proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_23.branch2to1_injector.ca_gamma - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_23.branch2to1_injector.cffn_gamma - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_23.branch2to1_injector.query_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_23.branch2to1_injector.query_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_23.branch2to1_injector.feat_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_23.branch2to1_injector.feat_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_23.branch2to1_injector.attn.sampling_offsets.weight - torch.Size([128, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_23.branch2to1_injector.attn.sampling_offsets.bias - torch.Size([128]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_23.branch2to1_injector.attn.attention_weights.weight - torch.Size([64, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_23.branch2to1_injector.attn.attention_weights.bias - torch.Size([64]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_23.branch2to1_injector.attn.value_proj.weight - torch.Size([384, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_23.branch2to1_injector.attn.value_proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_23.branch2to1_injector.attn.output_proj.weight - torch.Size([768, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_23.branch2to1_injector.attn.output_proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_23.branch2to1_injector.ffn.fc1.weight - torch.Size([192, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_23.branch2to1_injector.ffn.fc1.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_23.branch2to1_injector.ffn.dwconv.dwconv.weight - torch.Size([192, 1, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_23.branch2to1_injector.ffn.dwconv.dwconv.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_23.branch2to1_injector.ffn.fc2.weight - torch.Size([768, 192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_23.branch2to1_injector.ffn.fc2.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_23.branch2to1_injector.ffn_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_23.branch2to1_injector.ffn_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_23.branch1to2_proj.weight - torch.Size([384, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_23.branch1to2_proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_23.branch1to2_injector.ca_gamma - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_23.branch1to2_injector.cffn_gamma - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_23.branch1to2_injector.query_norm.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_23.branch1to2_injector.query_norm.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_23.branch1to2_injector.feat_norm.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_23.branch1to2_injector.feat_norm.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_23.branch1to2_injector.attn.sampling_offsets.weight - torch.Size([128, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_23.branch1to2_injector.attn.sampling_offsets.bias - torch.Size([128]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_23.branch1to2_injector.attn.attention_weights.weight - torch.Size([64, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_23.branch1to2_injector.attn.attention_weights.bias - torch.Size([64]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_23.branch1to2_injector.attn.value_proj.weight - torch.Size([192, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_23.branch1to2_injector.attn.value_proj.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_23.branch1to2_injector.attn.output_proj.weight - torch.Size([384, 192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_23.branch1to2_injector.attn.output_proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_23.branch1to2_injector.ffn.fc1.weight - torch.Size([96, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_23.branch1to2_injector.ffn.fc1.bias - torch.Size([96]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_23.branch1to2_injector.ffn.dwconv.dwconv.weight - torch.Size([96, 1, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_23.branch1to2_injector.ffn.dwconv.dwconv.bias - torch.Size([96]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_23.branch1to2_injector.ffn.fc2.weight - torch.Size([384, 96]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_23.branch1to2_injector.ffn.fc2.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_23.branch1to2_injector.ffn_norm.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.1.interaction_units_23.branch1to2_injector.ffn_norm.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_12.branch2to1_proj.weight - torch.Size([1024, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_12.branch2to1_proj.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_12.branch2to1_injector.ca_gamma - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_12.branch2to1_injector.cffn_gamma - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_12.branch2to1_injector.query_norm.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_12.branch2to1_injector.query_norm.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_12.branch2to1_injector.feat_norm.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_12.branch2to1_injector.feat_norm.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_12.branch2to1_injector.attn.sampling_offsets.weight - torch.Size([128, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_12.branch2to1_injector.attn.sampling_offsets.bias - torch.Size([128]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_12.branch2to1_injector.attn.attention_weights.weight - torch.Size([64, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_12.branch2to1_injector.attn.attention_weights.bias - torch.Size([64]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_12.branch2to1_injector.attn.value_proj.weight - torch.Size([512, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_12.branch2to1_injector.attn.value_proj.bias - torch.Size([512]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_12.branch2to1_injector.attn.output_proj.weight - torch.Size([1024, 512]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_12.branch2to1_injector.attn.output_proj.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_12.branch2to1_injector.ffn.fc1.weight - torch.Size([256, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_12.branch2to1_injector.ffn.fc1.bias - torch.Size([256]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_12.branch2to1_injector.ffn.dwconv.dwconv.weight - torch.Size([256, 1, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_12.branch2to1_injector.ffn.dwconv.dwconv.bias - torch.Size([256]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_12.branch2to1_injector.ffn.fc2.weight - torch.Size([1024, 256]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_12.branch2to1_injector.ffn.fc2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_12.branch2to1_injector.ffn_norm.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_12.branch2to1_injector.ffn_norm.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_12.branch1to2_proj.weight - torch.Size([768, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_12.branch1to2_proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_12.branch1to2_injector.ca_gamma - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_12.branch1to2_injector.cffn_gamma - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_12.branch1to2_injector.query_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_12.branch1to2_injector.query_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_12.branch1to2_injector.feat_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_12.branch1to2_injector.feat_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_12.branch1to2_injector.attn.sampling_offsets.weight - torch.Size([128, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_12.branch1to2_injector.attn.sampling_offsets.bias - torch.Size([128]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_12.branch1to2_injector.attn.attention_weights.weight - torch.Size([64, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_12.branch1to2_injector.attn.attention_weights.bias - torch.Size([64]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_12.branch1to2_injector.attn.value_proj.weight - torch.Size([384, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_12.branch1to2_injector.attn.value_proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_12.branch1to2_injector.attn.output_proj.weight - torch.Size([768, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_12.branch1to2_injector.attn.output_proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_12.branch1to2_injector.ffn.fc1.weight - torch.Size([192, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_12.branch1to2_injector.ffn.fc1.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_12.branch1to2_injector.ffn.dwconv.dwconv.weight - torch.Size([192, 1, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_12.branch1to2_injector.ffn.dwconv.dwconv.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_12.branch1to2_injector.ffn.fc2.weight - torch.Size([768, 192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_12.branch1to2_injector.ffn.fc2.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_12.branch1to2_injector.ffn_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_12.branch1to2_injector.ffn_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_23.branch2to1_proj.weight - torch.Size([768, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_23.branch2to1_proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_23.branch2to1_injector.ca_gamma - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_23.branch2to1_injector.cffn_gamma - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_23.branch2to1_injector.query_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_23.branch2to1_injector.query_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_23.branch2to1_injector.feat_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_23.branch2to1_injector.feat_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_23.branch2to1_injector.attn.sampling_offsets.weight - torch.Size([128, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_23.branch2to1_injector.attn.sampling_offsets.bias - torch.Size([128]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_23.branch2to1_injector.attn.attention_weights.weight - torch.Size([64, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_23.branch2to1_injector.attn.attention_weights.bias - torch.Size([64]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_23.branch2to1_injector.attn.value_proj.weight - torch.Size([384, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_23.branch2to1_injector.attn.value_proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_23.branch2to1_injector.attn.output_proj.weight - torch.Size([768, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_23.branch2to1_injector.attn.output_proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_23.branch2to1_injector.ffn.fc1.weight - torch.Size([192, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_23.branch2to1_injector.ffn.fc1.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_23.branch2to1_injector.ffn.dwconv.dwconv.weight - torch.Size([192, 1, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_23.branch2to1_injector.ffn.dwconv.dwconv.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_23.branch2to1_injector.ffn.fc2.weight - torch.Size([768, 192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_23.branch2to1_injector.ffn.fc2.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_23.branch2to1_injector.ffn_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_23.branch2to1_injector.ffn_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_23.branch1to2_proj.weight - torch.Size([384, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_23.branch1to2_proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_23.branch1to2_injector.ca_gamma - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_23.branch1to2_injector.cffn_gamma - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_23.branch1to2_injector.query_norm.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_23.branch1to2_injector.query_norm.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_23.branch1to2_injector.feat_norm.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_23.branch1to2_injector.feat_norm.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_23.branch1to2_injector.attn.sampling_offsets.weight - torch.Size([128, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_23.branch1to2_injector.attn.sampling_offsets.bias - torch.Size([128]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_23.branch1to2_injector.attn.attention_weights.weight - torch.Size([64, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_23.branch1to2_injector.attn.attention_weights.bias - torch.Size([64]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_23.branch1to2_injector.attn.value_proj.weight - torch.Size([192, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_23.branch1to2_injector.attn.value_proj.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_23.branch1to2_injector.attn.output_proj.weight - torch.Size([384, 192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_23.branch1to2_injector.attn.output_proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_23.branch1to2_injector.ffn.fc1.weight - torch.Size([96, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_23.branch1to2_injector.ffn.fc1.bias - torch.Size([96]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_23.branch1to2_injector.ffn.dwconv.dwconv.weight - torch.Size([96, 1, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_23.branch1to2_injector.ffn.dwconv.dwconv.bias - torch.Size([96]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_23.branch1to2_injector.ffn.fc2.weight - torch.Size([384, 96]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_23.branch1to2_injector.ffn.fc2.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_23.branch1to2_injector.ffn_norm.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.2.interaction_units_23.branch1to2_injector.ffn_norm.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_12.branch2to1_proj.weight - torch.Size([1024, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_12.branch2to1_proj.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_12.branch2to1_injector.ca_gamma - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_12.branch2to1_injector.cffn_gamma - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_12.branch2to1_injector.query_norm.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_12.branch2to1_injector.query_norm.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_12.branch2to1_injector.feat_norm.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_12.branch2to1_injector.feat_norm.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_12.branch2to1_injector.attn.sampling_offsets.weight - torch.Size([128, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_12.branch2to1_injector.attn.sampling_offsets.bias - torch.Size([128]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_12.branch2to1_injector.attn.attention_weights.weight - torch.Size([64, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_12.branch2to1_injector.attn.attention_weights.bias - torch.Size([64]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_12.branch2to1_injector.attn.value_proj.weight - torch.Size([512, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_12.branch2to1_injector.attn.value_proj.bias - torch.Size([512]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_12.branch2to1_injector.attn.output_proj.weight - torch.Size([1024, 512]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_12.branch2to1_injector.attn.output_proj.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_12.branch2to1_injector.ffn.fc1.weight - torch.Size([256, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_12.branch2to1_injector.ffn.fc1.bias - torch.Size([256]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_12.branch2to1_injector.ffn.dwconv.dwconv.weight - torch.Size([256, 1, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_12.branch2to1_injector.ffn.dwconv.dwconv.bias - torch.Size([256]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_12.branch2to1_injector.ffn.fc2.weight - torch.Size([1024, 256]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_12.branch2to1_injector.ffn.fc2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_12.branch2to1_injector.ffn_norm.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_12.branch2to1_injector.ffn_norm.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_12.branch1to2_proj.weight - torch.Size([768, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_12.branch1to2_proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_12.branch1to2_injector.ca_gamma - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_12.branch1to2_injector.cffn_gamma - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_12.branch1to2_injector.query_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_12.branch1to2_injector.query_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_12.branch1to2_injector.feat_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_12.branch1to2_injector.feat_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_12.branch1to2_injector.attn.sampling_offsets.weight - torch.Size([128, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_12.branch1to2_injector.attn.sampling_offsets.bias - torch.Size([128]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_12.branch1to2_injector.attn.attention_weights.weight - torch.Size([64, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_12.branch1to2_injector.attn.attention_weights.bias - torch.Size([64]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_12.branch1to2_injector.attn.value_proj.weight - torch.Size([384, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_12.branch1to2_injector.attn.value_proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_12.branch1to2_injector.attn.output_proj.weight - torch.Size([768, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_12.branch1to2_injector.attn.output_proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_12.branch1to2_injector.ffn.fc1.weight - torch.Size([192, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_12.branch1to2_injector.ffn.fc1.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_12.branch1to2_injector.ffn.dwconv.dwconv.weight - torch.Size([192, 1, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_12.branch1to2_injector.ffn.dwconv.dwconv.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_12.branch1to2_injector.ffn.fc2.weight - torch.Size([768, 192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_12.branch1to2_injector.ffn.fc2.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_12.branch1to2_injector.ffn_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_12.branch1to2_injector.ffn_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_23.branch2to1_proj.weight - torch.Size([768, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_23.branch2to1_proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_23.branch2to1_injector.ca_gamma - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_23.branch2to1_injector.cffn_gamma - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_23.branch2to1_injector.query_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_23.branch2to1_injector.query_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_23.branch2to1_injector.feat_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_23.branch2to1_injector.feat_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_23.branch2to1_injector.attn.sampling_offsets.weight - torch.Size([128, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_23.branch2to1_injector.attn.sampling_offsets.bias - torch.Size([128]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_23.branch2to1_injector.attn.attention_weights.weight - torch.Size([64, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_23.branch2to1_injector.attn.attention_weights.bias - torch.Size([64]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_23.branch2to1_injector.attn.value_proj.weight - torch.Size([384, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_23.branch2to1_injector.attn.value_proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_23.branch2to1_injector.attn.output_proj.weight - torch.Size([768, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_23.branch2to1_injector.attn.output_proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_23.branch2to1_injector.ffn.fc1.weight - torch.Size([192, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_23.branch2to1_injector.ffn.fc1.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_23.branch2to1_injector.ffn.dwconv.dwconv.weight - torch.Size([192, 1, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_23.branch2to1_injector.ffn.dwconv.dwconv.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_23.branch2to1_injector.ffn.fc2.weight - torch.Size([768, 192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_23.branch2to1_injector.ffn.fc2.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_23.branch2to1_injector.ffn_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_23.branch2to1_injector.ffn_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_23.branch1to2_proj.weight - torch.Size([384, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_23.branch1to2_proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_23.branch1to2_injector.ca_gamma - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_23.branch1to2_injector.cffn_gamma - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_23.branch1to2_injector.query_norm.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_23.branch1to2_injector.query_norm.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_23.branch1to2_injector.feat_norm.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_23.branch1to2_injector.feat_norm.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_23.branch1to2_injector.attn.sampling_offsets.weight - torch.Size([128, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_23.branch1to2_injector.attn.sampling_offsets.bias - torch.Size([128]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_23.branch1to2_injector.attn.attention_weights.weight - torch.Size([64, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_23.branch1to2_injector.attn.attention_weights.bias - torch.Size([64]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_23.branch1to2_injector.attn.value_proj.weight - torch.Size([192, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_23.branch1to2_injector.attn.value_proj.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_23.branch1to2_injector.attn.output_proj.weight - torch.Size([384, 192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_23.branch1to2_injector.attn.output_proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_23.branch1to2_injector.ffn.fc1.weight - torch.Size([96, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_23.branch1to2_injector.ffn.fc1.bias - torch.Size([96]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_23.branch1to2_injector.ffn.dwconv.dwconv.weight - torch.Size([96, 1, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_23.branch1to2_injector.ffn.dwconv.dwconv.bias - torch.Size([96]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_23.branch1to2_injector.ffn.fc2.weight - torch.Size([384, 96]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_23.branch1to2_injector.ffn.fc2.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_23.branch1to2_injector.ffn_norm.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.3.interaction_units_23.branch1to2_injector.ffn_norm.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_12.branch2to1_proj.weight - torch.Size([1024, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_12.branch2to1_proj.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_12.branch2to1_injector.ca_gamma - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_12.branch2to1_injector.cffn_gamma - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_12.branch2to1_injector.query_norm.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_12.branch2to1_injector.query_norm.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_12.branch2to1_injector.feat_norm.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_12.branch2to1_injector.feat_norm.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_12.branch2to1_injector.attn.sampling_offsets.weight - torch.Size([128, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_12.branch2to1_injector.attn.sampling_offsets.bias - torch.Size([128]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_12.branch2to1_injector.attn.attention_weights.weight - torch.Size([64, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_12.branch2to1_injector.attn.attention_weights.bias - torch.Size([64]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_12.branch2to1_injector.attn.value_proj.weight - torch.Size([512, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_12.branch2to1_injector.attn.value_proj.bias - torch.Size([512]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_12.branch2to1_injector.attn.output_proj.weight - torch.Size([1024, 512]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_12.branch2to1_injector.attn.output_proj.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_12.branch2to1_injector.ffn.fc1.weight - torch.Size([256, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_12.branch2to1_injector.ffn.fc1.bias - torch.Size([256]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_12.branch2to1_injector.ffn.dwconv.dwconv.weight - torch.Size([256, 1, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_12.branch2to1_injector.ffn.dwconv.dwconv.bias - torch.Size([256]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_12.branch2to1_injector.ffn.fc2.weight - torch.Size([1024, 256]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_12.branch2to1_injector.ffn.fc2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_12.branch2to1_injector.ffn_norm.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_12.branch2to1_injector.ffn_norm.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_12.branch1to2_proj.weight - torch.Size([768, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_12.branch1to2_proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_12.branch1to2_injector.ca_gamma - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_12.branch1to2_injector.cffn_gamma - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_12.branch1to2_injector.query_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_12.branch1to2_injector.query_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_12.branch1to2_injector.feat_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_12.branch1to2_injector.feat_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_12.branch1to2_injector.attn.sampling_offsets.weight - torch.Size([128, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_12.branch1to2_injector.attn.sampling_offsets.bias - torch.Size([128]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_12.branch1to2_injector.attn.attention_weights.weight - torch.Size([64, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_12.branch1to2_injector.attn.attention_weights.bias - torch.Size([64]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_12.branch1to2_injector.attn.value_proj.weight - torch.Size([384, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_12.branch1to2_injector.attn.value_proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_12.branch1to2_injector.attn.output_proj.weight - torch.Size([768, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_12.branch1to2_injector.attn.output_proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_12.branch1to2_injector.ffn.fc1.weight - torch.Size([192, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_12.branch1to2_injector.ffn.fc1.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_12.branch1to2_injector.ffn.dwconv.dwconv.weight - torch.Size([192, 1, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_12.branch1to2_injector.ffn.dwconv.dwconv.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_12.branch1to2_injector.ffn.fc2.weight - torch.Size([768, 192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_12.branch1to2_injector.ffn.fc2.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_12.branch1to2_injector.ffn_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_12.branch1to2_injector.ffn_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_23.branch2to1_proj.weight - torch.Size([768, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_23.branch2to1_proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_23.branch2to1_injector.ca_gamma - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_23.branch2to1_injector.cffn_gamma - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_23.branch2to1_injector.query_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_23.branch2to1_injector.query_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_23.branch2to1_injector.feat_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_23.branch2to1_injector.feat_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_23.branch2to1_injector.attn.sampling_offsets.weight - torch.Size([128, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_23.branch2to1_injector.attn.sampling_offsets.bias - torch.Size([128]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_23.branch2to1_injector.attn.attention_weights.weight - torch.Size([64, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_23.branch2to1_injector.attn.attention_weights.bias - torch.Size([64]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_23.branch2to1_injector.attn.value_proj.weight - torch.Size([384, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_23.branch2to1_injector.attn.value_proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_23.branch2to1_injector.attn.output_proj.weight - torch.Size([768, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_23.branch2to1_injector.attn.output_proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_23.branch2to1_injector.ffn.fc1.weight - torch.Size([192, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_23.branch2to1_injector.ffn.fc1.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_23.branch2to1_injector.ffn.dwconv.dwconv.weight - torch.Size([192, 1, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_23.branch2to1_injector.ffn.dwconv.dwconv.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_23.branch2to1_injector.ffn.fc2.weight - torch.Size([768, 192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_23.branch2to1_injector.ffn.fc2.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_23.branch2to1_injector.ffn_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_23.branch2to1_injector.ffn_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_23.branch1to2_proj.weight - torch.Size([384, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_23.branch1to2_proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_23.branch1to2_injector.ca_gamma - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_23.branch1to2_injector.cffn_gamma - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_23.branch1to2_injector.query_norm.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_23.branch1to2_injector.query_norm.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_23.branch1to2_injector.feat_norm.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_23.branch1to2_injector.feat_norm.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_23.branch1to2_injector.attn.sampling_offsets.weight - torch.Size([128, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_23.branch1to2_injector.attn.sampling_offsets.bias - torch.Size([128]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_23.branch1to2_injector.attn.attention_weights.weight - torch.Size([64, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_23.branch1to2_injector.attn.attention_weights.bias - torch.Size([64]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_23.branch1to2_injector.attn.value_proj.weight - torch.Size([192, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_23.branch1to2_injector.attn.value_proj.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_23.branch1to2_injector.attn.output_proj.weight - torch.Size([384, 192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_23.branch1to2_injector.attn.output_proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_23.branch1to2_injector.ffn.fc1.weight - torch.Size([96, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_23.branch1to2_injector.ffn.fc1.bias - torch.Size([96]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_23.branch1to2_injector.ffn.dwconv.dwconv.weight - torch.Size([96, 1, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_23.branch1to2_injector.ffn.dwconv.dwconv.bias - torch.Size([96]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_23.branch1to2_injector.ffn.fc2.weight - torch.Size([384, 96]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_23.branch1to2_injector.ffn.fc2.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_23.branch1to2_injector.ffn_norm.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.4.interaction_units_23.branch1to2_injector.ffn_norm.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_12.branch2to1_proj.weight - torch.Size([1024, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_12.branch2to1_proj.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_12.branch2to1_injector.ca_gamma - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_12.branch2to1_injector.cffn_gamma - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_12.branch2to1_injector.query_norm.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_12.branch2to1_injector.query_norm.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_12.branch2to1_injector.feat_norm.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_12.branch2to1_injector.feat_norm.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_12.branch2to1_injector.attn.sampling_offsets.weight - torch.Size([128, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_12.branch2to1_injector.attn.sampling_offsets.bias - torch.Size([128]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_12.branch2to1_injector.attn.attention_weights.weight - torch.Size([64, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_12.branch2to1_injector.attn.attention_weights.bias - torch.Size([64]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_12.branch2to1_injector.attn.value_proj.weight - torch.Size([512, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_12.branch2to1_injector.attn.value_proj.bias - torch.Size([512]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_12.branch2to1_injector.attn.output_proj.weight - torch.Size([1024, 512]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_12.branch2to1_injector.attn.output_proj.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_12.branch2to1_injector.ffn.fc1.weight - torch.Size([256, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_12.branch2to1_injector.ffn.fc1.bias - torch.Size([256]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_12.branch2to1_injector.ffn.dwconv.dwconv.weight - torch.Size([256, 1, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_12.branch2to1_injector.ffn.dwconv.dwconv.bias - torch.Size([256]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_12.branch2to1_injector.ffn.fc2.weight - torch.Size([1024, 256]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_12.branch2to1_injector.ffn.fc2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_12.branch2to1_injector.ffn_norm.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_12.branch2to1_injector.ffn_norm.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_12.branch1to2_proj.weight - torch.Size([768, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_12.branch1to2_proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_12.branch1to2_injector.ca_gamma - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_12.branch1to2_injector.cffn_gamma - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_12.branch1to2_injector.query_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_12.branch1to2_injector.query_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_12.branch1to2_injector.feat_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_12.branch1to2_injector.feat_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_12.branch1to2_injector.attn.sampling_offsets.weight - torch.Size([128, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_12.branch1to2_injector.attn.sampling_offsets.bias - torch.Size([128]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_12.branch1to2_injector.attn.attention_weights.weight - torch.Size([64, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_12.branch1to2_injector.attn.attention_weights.bias - torch.Size([64]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_12.branch1to2_injector.attn.value_proj.weight - torch.Size([384, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_12.branch1to2_injector.attn.value_proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_12.branch1to2_injector.attn.output_proj.weight - torch.Size([768, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_12.branch1to2_injector.attn.output_proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_12.branch1to2_injector.ffn.fc1.weight - torch.Size([192, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_12.branch1to2_injector.ffn.fc1.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_12.branch1to2_injector.ffn.dwconv.dwconv.weight - torch.Size([192, 1, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_12.branch1to2_injector.ffn.dwconv.dwconv.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_12.branch1to2_injector.ffn.fc2.weight - torch.Size([768, 192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_12.branch1to2_injector.ffn.fc2.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_12.branch1to2_injector.ffn_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_12.branch1to2_injector.ffn_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_23.branch2to1_proj.weight - torch.Size([768, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_23.branch2to1_proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_23.branch2to1_injector.ca_gamma - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_23.branch2to1_injector.cffn_gamma - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_23.branch2to1_injector.query_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_23.branch2to1_injector.query_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_23.branch2to1_injector.feat_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_23.branch2to1_injector.feat_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_23.branch2to1_injector.attn.sampling_offsets.weight - torch.Size([128, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_23.branch2to1_injector.attn.sampling_offsets.bias - torch.Size([128]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_23.branch2to1_injector.attn.attention_weights.weight - torch.Size([64, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_23.branch2to1_injector.attn.attention_weights.bias - torch.Size([64]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_23.branch2to1_injector.attn.value_proj.weight - torch.Size([384, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_23.branch2to1_injector.attn.value_proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_23.branch2to1_injector.attn.output_proj.weight - torch.Size([768, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_23.branch2to1_injector.attn.output_proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_23.branch2to1_injector.ffn.fc1.weight - torch.Size([192, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_23.branch2to1_injector.ffn.fc1.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_23.branch2to1_injector.ffn.dwconv.dwconv.weight - torch.Size([192, 1, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_23.branch2to1_injector.ffn.dwconv.dwconv.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_23.branch2to1_injector.ffn.fc2.weight - torch.Size([768, 192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_23.branch2to1_injector.ffn.fc2.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_23.branch2to1_injector.ffn_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_23.branch2to1_injector.ffn_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_23.branch1to2_proj.weight - torch.Size([384, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_23.branch1to2_proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_23.branch1to2_injector.ca_gamma - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_23.branch1to2_injector.cffn_gamma - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_23.branch1to2_injector.query_norm.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_23.branch1to2_injector.query_norm.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_23.branch1to2_injector.feat_norm.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_23.branch1to2_injector.feat_norm.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_23.branch1to2_injector.attn.sampling_offsets.weight - torch.Size([128, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_23.branch1to2_injector.attn.sampling_offsets.bias - torch.Size([128]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_23.branch1to2_injector.attn.attention_weights.weight - torch.Size([64, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_23.branch1to2_injector.attn.attention_weights.bias - torch.Size([64]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_23.branch1to2_injector.attn.value_proj.weight - torch.Size([192, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_23.branch1to2_injector.attn.value_proj.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_23.branch1to2_injector.attn.output_proj.weight - torch.Size([384, 192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_23.branch1to2_injector.attn.output_proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_23.branch1to2_injector.ffn.fc1.weight - torch.Size([96, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_23.branch1to2_injector.ffn.fc1.bias - torch.Size([96]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_23.branch1to2_injector.ffn.dwconv.dwconv.weight - torch.Size([96, 1, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_23.branch1to2_injector.ffn.dwconv.dwconv.bias - torch.Size([96]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_23.branch1to2_injector.ffn.fc2.weight - torch.Size([384, 96]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_23.branch1to2_injector.ffn.fc2.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_23.branch1to2_injector.ffn_norm.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.5.interaction_units_23.branch1to2_injector.ffn_norm.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_12.branch2to1_proj.weight - torch.Size([1024, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_12.branch2to1_proj.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_12.branch2to1_injector.ca_gamma - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_12.branch2to1_injector.cffn_gamma - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_12.branch2to1_injector.query_norm.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_12.branch2to1_injector.query_norm.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_12.branch2to1_injector.feat_norm.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_12.branch2to1_injector.feat_norm.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_12.branch2to1_injector.attn.sampling_offsets.weight - torch.Size([128, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_12.branch2to1_injector.attn.sampling_offsets.bias - torch.Size([128]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_12.branch2to1_injector.attn.attention_weights.weight - torch.Size([64, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_12.branch2to1_injector.attn.attention_weights.bias - torch.Size([64]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_12.branch2to1_injector.attn.value_proj.weight - torch.Size([512, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_12.branch2to1_injector.attn.value_proj.bias - torch.Size([512]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_12.branch2to1_injector.attn.output_proj.weight - torch.Size([1024, 512]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_12.branch2to1_injector.attn.output_proj.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_12.branch2to1_injector.ffn.fc1.weight - torch.Size([256, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_12.branch2to1_injector.ffn.fc1.bias - torch.Size([256]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_12.branch2to1_injector.ffn.dwconv.dwconv.weight - torch.Size([256, 1, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_12.branch2to1_injector.ffn.dwconv.dwconv.bias - torch.Size([256]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_12.branch2to1_injector.ffn.fc2.weight - torch.Size([1024, 256]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_12.branch2to1_injector.ffn.fc2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_12.branch2to1_injector.ffn_norm.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_12.branch2to1_injector.ffn_norm.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_12.branch1to2_proj.weight - torch.Size([768, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_12.branch1to2_proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_12.branch1to2_injector.ca_gamma - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_12.branch1to2_injector.cffn_gamma - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_12.branch1to2_injector.query_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_12.branch1to2_injector.query_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_12.branch1to2_injector.feat_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_12.branch1to2_injector.feat_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_12.branch1to2_injector.attn.sampling_offsets.weight - torch.Size([128, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_12.branch1to2_injector.attn.sampling_offsets.bias - torch.Size([128]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_12.branch1to2_injector.attn.attention_weights.weight - torch.Size([64, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_12.branch1to2_injector.attn.attention_weights.bias - torch.Size([64]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_12.branch1to2_injector.attn.value_proj.weight - torch.Size([384, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_12.branch1to2_injector.attn.value_proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_12.branch1to2_injector.attn.output_proj.weight - torch.Size([768, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_12.branch1to2_injector.attn.output_proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_12.branch1to2_injector.ffn.fc1.weight - torch.Size([192, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_12.branch1to2_injector.ffn.fc1.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_12.branch1to2_injector.ffn.dwconv.dwconv.weight - torch.Size([192, 1, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_12.branch1to2_injector.ffn.dwconv.dwconv.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_12.branch1to2_injector.ffn.fc2.weight - torch.Size([768, 192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_12.branch1to2_injector.ffn.fc2.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_12.branch1to2_injector.ffn_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_12.branch1to2_injector.ffn_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_23.branch2to1_proj.weight - torch.Size([768, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_23.branch2to1_proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_23.branch2to1_injector.ca_gamma - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_23.branch2to1_injector.cffn_gamma - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_23.branch2to1_injector.query_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_23.branch2to1_injector.query_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_23.branch2to1_injector.feat_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_23.branch2to1_injector.feat_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_23.branch2to1_injector.attn.sampling_offsets.weight - torch.Size([128, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_23.branch2to1_injector.attn.sampling_offsets.bias - torch.Size([128]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_23.branch2to1_injector.attn.attention_weights.weight - torch.Size([64, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_23.branch2to1_injector.attn.attention_weights.bias - torch.Size([64]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_23.branch2to1_injector.attn.value_proj.weight - torch.Size([384, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_23.branch2to1_injector.attn.value_proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_23.branch2to1_injector.attn.output_proj.weight - torch.Size([768, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_23.branch2to1_injector.attn.output_proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_23.branch2to1_injector.ffn.fc1.weight - torch.Size([192, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_23.branch2to1_injector.ffn.fc1.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_23.branch2to1_injector.ffn.dwconv.dwconv.weight - torch.Size([192, 1, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_23.branch2to1_injector.ffn.dwconv.dwconv.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_23.branch2to1_injector.ffn.fc2.weight - torch.Size([768, 192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_23.branch2to1_injector.ffn.fc2.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_23.branch2to1_injector.ffn_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_23.branch2to1_injector.ffn_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_23.branch1to2_proj.weight - torch.Size([384, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_23.branch1to2_proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_23.branch1to2_injector.ca_gamma - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_23.branch1to2_injector.cffn_gamma - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_23.branch1to2_injector.query_norm.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_23.branch1to2_injector.query_norm.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_23.branch1to2_injector.feat_norm.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_23.branch1to2_injector.feat_norm.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_23.branch1to2_injector.attn.sampling_offsets.weight - torch.Size([128, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_23.branch1to2_injector.attn.sampling_offsets.bias - torch.Size([128]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_23.branch1to2_injector.attn.attention_weights.weight - torch.Size([64, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_23.branch1to2_injector.attn.attention_weights.bias - torch.Size([64]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_23.branch1to2_injector.attn.value_proj.weight - torch.Size([192, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_23.branch1to2_injector.attn.value_proj.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_23.branch1to2_injector.attn.output_proj.weight - torch.Size([384, 192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_23.branch1to2_injector.attn.output_proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_23.branch1to2_injector.ffn.fc1.weight - torch.Size([96, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_23.branch1to2_injector.ffn.fc1.bias - torch.Size([96]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_23.branch1to2_injector.ffn.dwconv.dwconv.weight - torch.Size([96, 1, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_23.branch1to2_injector.ffn.dwconv.dwconv.bias - torch.Size([96]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_23.branch1to2_injector.ffn.fc2.weight - torch.Size([384, 96]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_23.branch1to2_injector.ffn.fc2.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_23.branch1to2_injector.ffn_norm.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.6.interaction_units_23.branch1to2_injector.ffn_norm.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_12.branch2to1_proj.weight - torch.Size([1024, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_12.branch2to1_proj.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_12.branch2to1_injector.ca_gamma - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_12.branch2to1_injector.cffn_gamma - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_12.branch2to1_injector.query_norm.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_12.branch2to1_injector.query_norm.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_12.branch2to1_injector.feat_norm.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_12.branch2to1_injector.feat_norm.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_12.branch2to1_injector.attn.sampling_offsets.weight - torch.Size([128, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_12.branch2to1_injector.attn.sampling_offsets.bias - torch.Size([128]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_12.branch2to1_injector.attn.attention_weights.weight - torch.Size([64, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_12.branch2to1_injector.attn.attention_weights.bias - torch.Size([64]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_12.branch2to1_injector.attn.value_proj.weight - torch.Size([512, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_12.branch2to1_injector.attn.value_proj.bias - torch.Size([512]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_12.branch2to1_injector.attn.output_proj.weight - torch.Size([1024, 512]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_12.branch2to1_injector.attn.output_proj.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_12.branch2to1_injector.ffn.fc1.weight - torch.Size([256, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_12.branch2to1_injector.ffn.fc1.bias - torch.Size([256]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_12.branch2to1_injector.ffn.dwconv.dwconv.weight - torch.Size([256, 1, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_12.branch2to1_injector.ffn.dwconv.dwconv.bias - torch.Size([256]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_12.branch2to1_injector.ffn.fc2.weight - torch.Size([1024, 256]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_12.branch2to1_injector.ffn.fc2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_12.branch2to1_injector.ffn_norm.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_12.branch2to1_injector.ffn_norm.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_12.branch1to2_proj.weight - torch.Size([768, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_12.branch1to2_proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_12.branch1to2_injector.ca_gamma - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_12.branch1to2_injector.cffn_gamma - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_12.branch1to2_injector.query_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_12.branch1to2_injector.query_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_12.branch1to2_injector.feat_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_12.branch1to2_injector.feat_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_12.branch1to2_injector.attn.sampling_offsets.weight - torch.Size([128, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_12.branch1to2_injector.attn.sampling_offsets.bias - torch.Size([128]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_12.branch1to2_injector.attn.attention_weights.weight - torch.Size([64, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_12.branch1to2_injector.attn.attention_weights.bias - torch.Size([64]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_12.branch1to2_injector.attn.value_proj.weight - torch.Size([384, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_12.branch1to2_injector.attn.value_proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_12.branch1to2_injector.attn.output_proj.weight - torch.Size([768, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_12.branch1to2_injector.attn.output_proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_12.branch1to2_injector.ffn.fc1.weight - torch.Size([192, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_12.branch1to2_injector.ffn.fc1.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_12.branch1to2_injector.ffn.dwconv.dwconv.weight - torch.Size([192, 1, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_12.branch1to2_injector.ffn.dwconv.dwconv.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_12.branch1to2_injector.ffn.fc2.weight - torch.Size([768, 192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_12.branch1to2_injector.ffn.fc2.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_12.branch1to2_injector.ffn_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_12.branch1to2_injector.ffn_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_23.branch2to1_proj.weight - torch.Size([768, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_23.branch2to1_proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_23.branch2to1_injector.ca_gamma - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_23.branch2to1_injector.cffn_gamma - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_23.branch2to1_injector.query_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_23.branch2to1_injector.query_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_23.branch2to1_injector.feat_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_23.branch2to1_injector.feat_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_23.branch2to1_injector.attn.sampling_offsets.weight - torch.Size([128, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_23.branch2to1_injector.attn.sampling_offsets.bias - torch.Size([128]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_23.branch2to1_injector.attn.attention_weights.weight - torch.Size([64, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_23.branch2to1_injector.attn.attention_weights.bias - torch.Size([64]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_23.branch2to1_injector.attn.value_proj.weight - torch.Size([384, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_23.branch2to1_injector.attn.value_proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_23.branch2to1_injector.attn.output_proj.weight - torch.Size([768, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_23.branch2to1_injector.attn.output_proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_23.branch2to1_injector.ffn.fc1.weight - torch.Size([192, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_23.branch2to1_injector.ffn.fc1.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_23.branch2to1_injector.ffn.dwconv.dwconv.weight - torch.Size([192, 1, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_23.branch2to1_injector.ffn.dwconv.dwconv.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_23.branch2to1_injector.ffn.fc2.weight - torch.Size([768, 192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_23.branch2to1_injector.ffn.fc2.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_23.branch2to1_injector.ffn_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_23.branch2to1_injector.ffn_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_23.branch1to2_proj.weight - torch.Size([384, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_23.branch1to2_proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_23.branch1to2_injector.ca_gamma - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_23.branch1to2_injector.cffn_gamma - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_23.branch1to2_injector.query_norm.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_23.branch1to2_injector.query_norm.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_23.branch1to2_injector.feat_norm.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_23.branch1to2_injector.feat_norm.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_23.branch1to2_injector.attn.sampling_offsets.weight - torch.Size([128, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_23.branch1to2_injector.attn.sampling_offsets.bias - torch.Size([128]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_23.branch1to2_injector.attn.attention_weights.weight - torch.Size([64, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_23.branch1to2_injector.attn.attention_weights.bias - torch.Size([64]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_23.branch1to2_injector.attn.value_proj.weight - torch.Size([192, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_23.branch1to2_injector.attn.value_proj.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_23.branch1to2_injector.attn.output_proj.weight - torch.Size([384, 192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_23.branch1to2_injector.attn.output_proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_23.branch1to2_injector.ffn.fc1.weight - torch.Size([96, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_23.branch1to2_injector.ffn.fc1.bias - torch.Size([96]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_23.branch1to2_injector.ffn.dwconv.dwconv.weight - torch.Size([96, 1, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_23.branch1to2_injector.ffn.dwconv.dwconv.bias - torch.Size([96]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_23.branch1to2_injector.ffn.fc2.weight - torch.Size([384, 96]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_23.branch1to2_injector.ffn.fc2.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_23.branch1to2_injector.ffn_norm.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.7.interaction_units_23.branch1to2_injector.ffn_norm.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_12.branch2to1_proj.weight - torch.Size([1024, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_12.branch2to1_proj.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_12.branch2to1_injector.ca_gamma - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_12.branch2to1_injector.cffn_gamma - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_12.branch2to1_injector.query_norm.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_12.branch2to1_injector.query_norm.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_12.branch2to1_injector.feat_norm.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_12.branch2to1_injector.feat_norm.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_12.branch2to1_injector.attn.sampling_offsets.weight - torch.Size([128, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_12.branch2to1_injector.attn.sampling_offsets.bias - torch.Size([128]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_12.branch2to1_injector.attn.attention_weights.weight - torch.Size([64, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_12.branch2to1_injector.attn.attention_weights.bias - torch.Size([64]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_12.branch2to1_injector.attn.value_proj.weight - torch.Size([512, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_12.branch2to1_injector.attn.value_proj.bias - torch.Size([512]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_12.branch2to1_injector.attn.output_proj.weight - torch.Size([1024, 512]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_12.branch2to1_injector.attn.output_proj.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_12.branch2to1_injector.ffn.fc1.weight - torch.Size([256, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_12.branch2to1_injector.ffn.fc1.bias - torch.Size([256]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_12.branch2to1_injector.ffn.dwconv.dwconv.weight - torch.Size([256, 1, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_12.branch2to1_injector.ffn.dwconv.dwconv.bias - torch.Size([256]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_12.branch2to1_injector.ffn.fc2.weight - torch.Size([1024, 256]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_12.branch2to1_injector.ffn.fc2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_12.branch2to1_injector.ffn_norm.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_12.branch2to1_injector.ffn_norm.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_12.branch1to2_proj.weight - torch.Size([768, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_12.branch1to2_proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_12.branch1to2_injector.ca_gamma - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_12.branch1to2_injector.cffn_gamma - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_12.branch1to2_injector.query_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_12.branch1to2_injector.query_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_12.branch1to2_injector.feat_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_12.branch1to2_injector.feat_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_12.branch1to2_injector.attn.sampling_offsets.weight - torch.Size([128, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_12.branch1to2_injector.attn.sampling_offsets.bias - torch.Size([128]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_12.branch1to2_injector.attn.attention_weights.weight - torch.Size([64, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_12.branch1to2_injector.attn.attention_weights.bias - torch.Size([64]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_12.branch1to2_injector.attn.value_proj.weight - torch.Size([384, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_12.branch1to2_injector.attn.value_proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_12.branch1to2_injector.attn.output_proj.weight - torch.Size([768, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_12.branch1to2_injector.attn.output_proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_12.branch1to2_injector.ffn.fc1.weight - torch.Size([192, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_12.branch1to2_injector.ffn.fc1.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_12.branch1to2_injector.ffn.dwconv.dwconv.weight - torch.Size([192, 1, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_12.branch1to2_injector.ffn.dwconv.dwconv.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_12.branch1to2_injector.ffn.fc2.weight - torch.Size([768, 192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_12.branch1to2_injector.ffn.fc2.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_12.branch1to2_injector.ffn_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_12.branch1to2_injector.ffn_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_23.branch2to1_proj.weight - torch.Size([768, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_23.branch2to1_proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_23.branch2to1_injector.ca_gamma - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_23.branch2to1_injector.cffn_gamma - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_23.branch2to1_injector.query_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_23.branch2to1_injector.query_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_23.branch2to1_injector.feat_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_23.branch2to1_injector.feat_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_23.branch2to1_injector.attn.sampling_offsets.weight - torch.Size([128, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_23.branch2to1_injector.attn.sampling_offsets.bias - torch.Size([128]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_23.branch2to1_injector.attn.attention_weights.weight - torch.Size([64, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_23.branch2to1_injector.attn.attention_weights.bias - torch.Size([64]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_23.branch2to1_injector.attn.value_proj.weight - torch.Size([384, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_23.branch2to1_injector.attn.value_proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_23.branch2to1_injector.attn.output_proj.weight - torch.Size([768, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_23.branch2to1_injector.attn.output_proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_23.branch2to1_injector.ffn.fc1.weight - torch.Size([192, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_23.branch2to1_injector.ffn.fc1.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_23.branch2to1_injector.ffn.dwconv.dwconv.weight - torch.Size([192, 1, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_23.branch2to1_injector.ffn.dwconv.dwconv.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_23.branch2to1_injector.ffn.fc2.weight - torch.Size([768, 192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_23.branch2to1_injector.ffn.fc2.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_23.branch2to1_injector.ffn_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_23.branch2to1_injector.ffn_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_23.branch1to2_proj.weight - torch.Size([384, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_23.branch1to2_proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_23.branch1to2_injector.ca_gamma - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_23.branch1to2_injector.cffn_gamma - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_23.branch1to2_injector.query_norm.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_23.branch1to2_injector.query_norm.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_23.branch1to2_injector.feat_norm.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_23.branch1to2_injector.feat_norm.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_23.branch1to2_injector.attn.sampling_offsets.weight - torch.Size([128, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_23.branch1to2_injector.attn.sampling_offsets.bias - torch.Size([128]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_23.branch1to2_injector.attn.attention_weights.weight - torch.Size([64, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_23.branch1to2_injector.attn.attention_weights.bias - torch.Size([64]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_23.branch1to2_injector.attn.value_proj.weight - torch.Size([192, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_23.branch1to2_injector.attn.value_proj.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_23.branch1to2_injector.attn.output_proj.weight - torch.Size([384, 192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_23.branch1to2_injector.attn.output_proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_23.branch1to2_injector.ffn.fc1.weight - torch.Size([96, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_23.branch1to2_injector.ffn.fc1.bias - torch.Size([96]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_23.branch1to2_injector.ffn.dwconv.dwconv.weight - torch.Size([96, 1, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_23.branch1to2_injector.ffn.dwconv.dwconv.bias - torch.Size([96]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_23.branch1to2_injector.ffn.fc2.weight - torch.Size([384, 96]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_23.branch1to2_injector.ffn.fc2.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_23.branch1to2_injector.ffn_norm.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.8.interaction_units_23.branch1to2_injector.ffn_norm.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_12.branch2to1_proj.weight - torch.Size([1024, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_12.branch2to1_proj.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_12.branch2to1_injector.ca_gamma - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_12.branch2to1_injector.cffn_gamma - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_12.branch2to1_injector.query_norm.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_12.branch2to1_injector.query_norm.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_12.branch2to1_injector.feat_norm.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_12.branch2to1_injector.feat_norm.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_12.branch2to1_injector.attn.sampling_offsets.weight - torch.Size([128, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_12.branch2to1_injector.attn.sampling_offsets.bias - torch.Size([128]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_12.branch2to1_injector.attn.attention_weights.weight - torch.Size([64, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_12.branch2to1_injector.attn.attention_weights.bias - torch.Size([64]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_12.branch2to1_injector.attn.value_proj.weight - torch.Size([512, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_12.branch2to1_injector.attn.value_proj.bias - torch.Size([512]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_12.branch2to1_injector.attn.output_proj.weight - torch.Size([1024, 512]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_12.branch2to1_injector.attn.output_proj.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_12.branch2to1_injector.ffn.fc1.weight - torch.Size([256, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_12.branch2to1_injector.ffn.fc1.bias - torch.Size([256]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_12.branch2to1_injector.ffn.dwconv.dwconv.weight - torch.Size([256, 1, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_12.branch2to1_injector.ffn.dwconv.dwconv.bias - torch.Size([256]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_12.branch2to1_injector.ffn.fc2.weight - torch.Size([1024, 256]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_12.branch2to1_injector.ffn.fc2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_12.branch2to1_injector.ffn_norm.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_12.branch2to1_injector.ffn_norm.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_12.branch1to2_proj.weight - torch.Size([768, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_12.branch1to2_proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_12.branch1to2_injector.ca_gamma - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_12.branch1to2_injector.cffn_gamma - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_12.branch1to2_injector.query_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_12.branch1to2_injector.query_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_12.branch1to2_injector.feat_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_12.branch1to2_injector.feat_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_12.branch1to2_injector.attn.sampling_offsets.weight - torch.Size([128, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_12.branch1to2_injector.attn.sampling_offsets.bias - torch.Size([128]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_12.branch1to2_injector.attn.attention_weights.weight - torch.Size([64, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_12.branch1to2_injector.attn.attention_weights.bias - torch.Size([64]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_12.branch1to2_injector.attn.value_proj.weight - torch.Size([384, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_12.branch1to2_injector.attn.value_proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_12.branch1to2_injector.attn.output_proj.weight - torch.Size([768, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_12.branch1to2_injector.attn.output_proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_12.branch1to2_injector.ffn.fc1.weight - torch.Size([192, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_12.branch1to2_injector.ffn.fc1.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_12.branch1to2_injector.ffn.dwconv.dwconv.weight - torch.Size([192, 1, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_12.branch1to2_injector.ffn.dwconv.dwconv.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_12.branch1to2_injector.ffn.fc2.weight - torch.Size([768, 192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_12.branch1to2_injector.ffn.fc2.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_12.branch1to2_injector.ffn_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_12.branch1to2_injector.ffn_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_23.branch2to1_proj.weight - torch.Size([768, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_23.branch2to1_proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_23.branch2to1_injector.ca_gamma - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_23.branch2to1_injector.cffn_gamma - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_23.branch2to1_injector.query_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_23.branch2to1_injector.query_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_23.branch2to1_injector.feat_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_23.branch2to1_injector.feat_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_23.branch2to1_injector.attn.sampling_offsets.weight - torch.Size([128, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_23.branch2to1_injector.attn.sampling_offsets.bias - torch.Size([128]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_23.branch2to1_injector.attn.attention_weights.weight - torch.Size([64, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_23.branch2to1_injector.attn.attention_weights.bias - torch.Size([64]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_23.branch2to1_injector.attn.value_proj.weight - torch.Size([384, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_23.branch2to1_injector.attn.value_proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_23.branch2to1_injector.attn.output_proj.weight - torch.Size([768, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_23.branch2to1_injector.attn.output_proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_23.branch2to1_injector.ffn.fc1.weight - torch.Size([192, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_23.branch2to1_injector.ffn.fc1.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_23.branch2to1_injector.ffn.dwconv.dwconv.weight - torch.Size([192, 1, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_23.branch2to1_injector.ffn.dwconv.dwconv.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_23.branch2to1_injector.ffn.fc2.weight - torch.Size([768, 192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_23.branch2to1_injector.ffn.fc2.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_23.branch2to1_injector.ffn_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_23.branch2to1_injector.ffn_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_23.branch1to2_proj.weight - torch.Size([384, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_23.branch1to2_proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_23.branch1to2_injector.ca_gamma - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_23.branch1to2_injector.cffn_gamma - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_23.branch1to2_injector.query_norm.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_23.branch1to2_injector.query_norm.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_23.branch1to2_injector.feat_norm.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_23.branch1to2_injector.feat_norm.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_23.branch1to2_injector.attn.sampling_offsets.weight - torch.Size([128, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_23.branch1to2_injector.attn.sampling_offsets.bias - torch.Size([128]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_23.branch1to2_injector.attn.attention_weights.weight - torch.Size([64, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_23.branch1to2_injector.attn.attention_weights.bias - torch.Size([64]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_23.branch1to2_injector.attn.value_proj.weight - torch.Size([192, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_23.branch1to2_injector.attn.value_proj.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_23.branch1to2_injector.attn.output_proj.weight - torch.Size([384, 192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_23.branch1to2_injector.attn.output_proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_23.branch1to2_injector.ffn.fc1.weight - torch.Size([96, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_23.branch1to2_injector.ffn.fc1.bias - torch.Size([96]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_23.branch1to2_injector.ffn.dwconv.dwconv.weight - torch.Size([96, 1, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_23.branch1to2_injector.ffn.dwconv.dwconv.bias - torch.Size([96]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_23.branch1to2_injector.ffn.fc2.weight - torch.Size([384, 96]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_23.branch1to2_injector.ffn.fc2.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_23.branch1to2_injector.ffn_norm.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.9.interaction_units_23.branch1to2_injector.ffn_norm.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_12.branch2to1_proj.weight - torch.Size([1024, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_12.branch2to1_proj.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_12.branch2to1_injector.ca_gamma - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_12.branch2to1_injector.cffn_gamma - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_12.branch2to1_injector.query_norm.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_12.branch2to1_injector.query_norm.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_12.branch2to1_injector.feat_norm.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_12.branch2to1_injector.feat_norm.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_12.branch2to1_injector.attn.sampling_offsets.weight - torch.Size([128, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_12.branch2to1_injector.attn.sampling_offsets.bias - torch.Size([128]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_12.branch2to1_injector.attn.attention_weights.weight - torch.Size([64, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_12.branch2to1_injector.attn.attention_weights.bias - torch.Size([64]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_12.branch2to1_injector.attn.value_proj.weight - torch.Size([512, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_12.branch2to1_injector.attn.value_proj.bias - torch.Size([512]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_12.branch2to1_injector.attn.output_proj.weight - torch.Size([1024, 512]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_12.branch2to1_injector.attn.output_proj.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_12.branch2to1_injector.ffn.fc1.weight - torch.Size([256, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_12.branch2to1_injector.ffn.fc1.bias - torch.Size([256]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_12.branch2to1_injector.ffn.dwconv.dwconv.weight - torch.Size([256, 1, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_12.branch2to1_injector.ffn.dwconv.dwconv.bias - torch.Size([256]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_12.branch2to1_injector.ffn.fc2.weight - torch.Size([1024, 256]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_12.branch2to1_injector.ffn.fc2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_12.branch2to1_injector.ffn_norm.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_12.branch2to1_injector.ffn_norm.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_12.branch1to2_proj.weight - torch.Size([768, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_12.branch1to2_proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_12.branch1to2_injector.ca_gamma - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_12.branch1to2_injector.cffn_gamma - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_12.branch1to2_injector.query_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_12.branch1to2_injector.query_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_12.branch1to2_injector.feat_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_12.branch1to2_injector.feat_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_12.branch1to2_injector.attn.sampling_offsets.weight - torch.Size([128, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_12.branch1to2_injector.attn.sampling_offsets.bias - torch.Size([128]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_12.branch1to2_injector.attn.attention_weights.weight - torch.Size([64, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_12.branch1to2_injector.attn.attention_weights.bias - torch.Size([64]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_12.branch1to2_injector.attn.value_proj.weight - torch.Size([384, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_12.branch1to2_injector.attn.value_proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_12.branch1to2_injector.attn.output_proj.weight - torch.Size([768, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_12.branch1to2_injector.attn.output_proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_12.branch1to2_injector.ffn.fc1.weight - torch.Size([192, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_12.branch1to2_injector.ffn.fc1.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_12.branch1to2_injector.ffn.dwconv.dwconv.weight - torch.Size([192, 1, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_12.branch1to2_injector.ffn.dwconv.dwconv.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_12.branch1to2_injector.ffn.fc2.weight - torch.Size([768, 192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_12.branch1to2_injector.ffn.fc2.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_12.branch1to2_injector.ffn_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_12.branch1to2_injector.ffn_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_23.branch2to1_proj.weight - torch.Size([768, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_23.branch2to1_proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_23.branch2to1_injector.ca_gamma - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_23.branch2to1_injector.cffn_gamma - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_23.branch2to1_injector.query_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_23.branch2to1_injector.query_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_23.branch2to1_injector.feat_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_23.branch2to1_injector.feat_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_23.branch2to1_injector.attn.sampling_offsets.weight - torch.Size([128, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_23.branch2to1_injector.attn.sampling_offsets.bias - torch.Size([128]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_23.branch2to1_injector.attn.attention_weights.weight - torch.Size([64, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_23.branch2to1_injector.attn.attention_weights.bias - torch.Size([64]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_23.branch2to1_injector.attn.value_proj.weight - torch.Size([384, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_23.branch2to1_injector.attn.value_proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_23.branch2to1_injector.attn.output_proj.weight - torch.Size([768, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_23.branch2to1_injector.attn.output_proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_23.branch2to1_injector.ffn.fc1.weight - torch.Size([192, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_23.branch2to1_injector.ffn.fc1.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_23.branch2to1_injector.ffn.dwconv.dwconv.weight - torch.Size([192, 1, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_23.branch2to1_injector.ffn.dwconv.dwconv.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_23.branch2to1_injector.ffn.fc2.weight - torch.Size([768, 192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_23.branch2to1_injector.ffn.fc2.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_23.branch2to1_injector.ffn_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_23.branch2to1_injector.ffn_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_23.branch1to2_proj.weight - torch.Size([384, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_23.branch1to2_proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_23.branch1to2_injector.ca_gamma - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_23.branch1to2_injector.cffn_gamma - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_23.branch1to2_injector.query_norm.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_23.branch1to2_injector.query_norm.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_23.branch1to2_injector.feat_norm.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_23.branch1to2_injector.feat_norm.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_23.branch1to2_injector.attn.sampling_offsets.weight - torch.Size([128, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_23.branch1to2_injector.attn.sampling_offsets.bias - torch.Size([128]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_23.branch1to2_injector.attn.attention_weights.weight - torch.Size([64, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_23.branch1to2_injector.attn.attention_weights.bias - torch.Size([64]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_23.branch1to2_injector.attn.value_proj.weight - torch.Size([192, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_23.branch1to2_injector.attn.value_proj.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_23.branch1to2_injector.attn.output_proj.weight - torch.Size([384, 192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_23.branch1to2_injector.attn.output_proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_23.branch1to2_injector.ffn.fc1.weight - torch.Size([96, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_23.branch1to2_injector.ffn.fc1.bias - torch.Size([96]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_23.branch1to2_injector.ffn.dwconv.dwconv.weight - torch.Size([96, 1, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_23.branch1to2_injector.ffn.dwconv.dwconv.bias - torch.Size([96]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_23.branch1to2_injector.ffn.fc2.weight - torch.Size([384, 96]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_23.branch1to2_injector.ffn.fc2.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_23.branch1to2_injector.ffn_norm.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.10.interaction_units_23.branch1to2_injector.ffn_norm.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_12.branch2to1_proj.weight - torch.Size([1024, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_12.branch2to1_proj.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_12.branch2to1_injector.ca_gamma - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_12.branch2to1_injector.cffn_gamma - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_12.branch2to1_injector.query_norm.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_12.branch2to1_injector.query_norm.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_12.branch2to1_injector.feat_norm.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_12.branch2to1_injector.feat_norm.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_12.branch2to1_injector.attn.sampling_offsets.weight - torch.Size([128, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_12.branch2to1_injector.attn.sampling_offsets.bias - torch.Size([128]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_12.branch2to1_injector.attn.attention_weights.weight - torch.Size([64, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_12.branch2to1_injector.attn.attention_weights.bias - torch.Size([64]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_12.branch2to1_injector.attn.value_proj.weight - torch.Size([512, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_12.branch2to1_injector.attn.value_proj.bias - torch.Size([512]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_12.branch2to1_injector.attn.output_proj.weight - torch.Size([1024, 512]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_12.branch2to1_injector.attn.output_proj.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_12.branch2to1_injector.ffn.fc1.weight - torch.Size([256, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_12.branch2to1_injector.ffn.fc1.bias - torch.Size([256]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_12.branch2to1_injector.ffn.dwconv.dwconv.weight - torch.Size([256, 1, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_12.branch2to1_injector.ffn.dwconv.dwconv.bias - torch.Size([256]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_12.branch2to1_injector.ffn.fc2.weight - torch.Size([1024, 256]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_12.branch2to1_injector.ffn.fc2.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_12.branch2to1_injector.ffn_norm.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_12.branch2to1_injector.ffn_norm.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_12.branch1to2_proj.weight - torch.Size([768, 1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_12.branch1to2_proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_12.branch1to2_injector.ca_gamma - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_12.branch1to2_injector.cffn_gamma - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_12.branch1to2_injector.query_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_12.branch1to2_injector.query_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_12.branch1to2_injector.feat_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_12.branch1to2_injector.feat_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_12.branch1to2_injector.attn.sampling_offsets.weight - torch.Size([128, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_12.branch1to2_injector.attn.sampling_offsets.bias - torch.Size([128]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_12.branch1to2_injector.attn.attention_weights.weight - torch.Size([64, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_12.branch1to2_injector.attn.attention_weights.bias - torch.Size([64]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_12.branch1to2_injector.attn.value_proj.weight - torch.Size([384, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_12.branch1to2_injector.attn.value_proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_12.branch1to2_injector.attn.output_proj.weight - torch.Size([768, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_12.branch1to2_injector.attn.output_proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_12.branch1to2_injector.ffn.fc1.weight - torch.Size([192, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_12.branch1to2_injector.ffn.fc1.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_12.branch1to2_injector.ffn.dwconv.dwconv.weight - torch.Size([192, 1, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_12.branch1to2_injector.ffn.dwconv.dwconv.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_12.branch1to2_injector.ffn.fc2.weight - torch.Size([768, 192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_12.branch1to2_injector.ffn.fc2.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_12.branch1to2_injector.ffn_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_12.branch1to2_injector.ffn_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_23.branch2to1_proj.weight - torch.Size([768, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_23.branch2to1_proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_23.branch2to1_injector.ca_gamma - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_23.branch2to1_injector.cffn_gamma - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_23.branch2to1_injector.query_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_23.branch2to1_injector.query_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_23.branch2to1_injector.feat_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_23.branch2to1_injector.feat_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_23.branch2to1_injector.attn.sampling_offsets.weight - torch.Size([128, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_23.branch2to1_injector.attn.sampling_offsets.bias - torch.Size([128]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_23.branch2to1_injector.attn.attention_weights.weight - torch.Size([64, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_23.branch2to1_injector.attn.attention_weights.bias - torch.Size([64]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_23.branch2to1_injector.attn.value_proj.weight - torch.Size([384, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_23.branch2to1_injector.attn.value_proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_23.branch2to1_injector.attn.output_proj.weight - torch.Size([768, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_23.branch2to1_injector.attn.output_proj.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_23.branch2to1_injector.ffn.fc1.weight - torch.Size([192, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_23.branch2to1_injector.ffn.fc1.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_23.branch2to1_injector.ffn.dwconv.dwconv.weight - torch.Size([192, 1, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_23.branch2to1_injector.ffn.dwconv.dwconv.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_23.branch2to1_injector.ffn.fc2.weight - torch.Size([768, 192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_23.branch2to1_injector.ffn.fc2.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_23.branch2to1_injector.ffn_norm.weight - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_23.branch2to1_injector.ffn_norm.bias - torch.Size([768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_23.branch1to2_proj.weight - torch.Size([384, 768]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_23.branch1to2_proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_23.branch1to2_injector.ca_gamma - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_23.branch1to2_injector.cffn_gamma - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_23.branch1to2_injector.query_norm.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_23.branch1to2_injector.query_norm.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_23.branch1to2_injector.feat_norm.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_23.branch1to2_injector.feat_norm.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_23.branch1to2_injector.attn.sampling_offsets.weight - torch.Size([128, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_23.branch1to2_injector.attn.sampling_offsets.bias - torch.Size([128]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_23.branch1to2_injector.attn.attention_weights.weight - torch.Size([64, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_23.branch1to2_injector.attn.attention_weights.bias - torch.Size([64]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_23.branch1to2_injector.attn.value_proj.weight - torch.Size([192, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_23.branch1to2_injector.attn.value_proj.bias - torch.Size([192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_23.branch1to2_injector.attn.output_proj.weight - torch.Size([384, 192]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_23.branch1to2_injector.attn.output_proj.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_23.branch1to2_injector.ffn.fc1.weight - torch.Size([96, 384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_23.branch1to2_injector.ffn.fc1.bias - torch.Size([96]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_23.branch1to2_injector.ffn.dwconv.dwconv.weight - torch.Size([96, 1, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_23.branch1to2_injector.ffn.dwconv.dwconv.bias - torch.Size([96]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_23.branch1to2_injector.ffn.fc2.weight - torch.Size([384, 96]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_23.branch1to2_injector.ffn.fc2.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_23.branch1to2_injector.ffn_norm.weight - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.interactions.11.interaction_units_23.branch1to2_injector.ffn_norm.bias - torch.Size([384]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.merge_branch1.0.weight - torch.Size([1024, 1024, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.merge_branch1.1.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.merge_branch1.1.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.merge_branch1.3.weight - torch.Size([1024, 1024, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.merge_branch1.4.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.merge_branch1.4.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.merge_branch2.0.weight - torch.Size([1024, 768, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.merge_branch2.1.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.merge_branch2.1.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.merge_branch2.3.weight - torch.Size([1024, 1024, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.merge_branch2.4.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.merge_branch2.4.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.merge_branch3.0.weight - torch.Size([1024, 384, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.merge_branch3.1.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.merge_branch3.1.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.merge_branch3.3.weight - torch.Size([1024, 1024, 3, 3]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.merge_branch3.4.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.merge_branch3.4.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.fpn1.0.weight - torch.Size([1024, 1024, 2, 2]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.fpn1.0.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.fpn1.1.weight - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.fpn1.1.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.fpn1.3.weight - torch.Size([1024, 1024, 2, 2]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.fpn1.3.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.fpn2.0.weight - torch.Size([1024, 1024, 2, 2]): +The value is the same before and after calling `init_weights` of MaskRCNN + +backbone.fpn2.0.bias - torch.Size([1024]): +The value is the same before and after calling `init_weights` of MaskRCNN + +neck.lateral_convs.0.conv.weight - torch.Size([256, 1024, 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 MaskRCNN + +neck.lateral_convs.1.conv.weight - torch.Size([256, 1024, 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 MaskRCNN + +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 MaskRCNN + +neck.lateral_convs.3.conv.weight - torch.Size([256, 1024, 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 MaskRCNN + +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 MaskRCNN + +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 MaskRCNN + +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 MaskRCNN + +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 MaskRCNN + +rpn_head.rpn_conv.weight - torch.Size([256, 256, 3, 3]): +NormalInit: mean=0, std=0.01, bias=0 + +rpn_head.rpn_conv.bias - torch.Size([256]): +NormalInit: mean=0, std=0.01, bias=0 + +rpn_head.rpn_cls.weight - torch.Size([3, 256, 1, 1]): +NormalInit: mean=0, std=0.01, bias=0 + +rpn_head.rpn_cls.bias - torch.Size([3]): +NormalInit: mean=0, std=0.01, bias=0 + +rpn_head.rpn_reg.weight - torch.Size([12, 256, 1, 1]): +NormalInit: mean=0, std=0.01, bias=0 + +rpn_head.rpn_reg.bias - torch.Size([12]): +NormalInit: mean=0, std=0.01, bias=0 + +roi_head.bbox_head.fc_cls.weight - torch.Size([81, 1024]): +NormalInit: mean=0, std=0.01, bias=0 + +roi_head.bbox_head.fc_cls.bias - torch.Size([81]): +NormalInit: mean=0, std=0.01, bias=0 + +roi_head.bbox_head.fc_reg.weight - torch.Size([320, 1024]): +NormalInit: mean=0, std=0.001, bias=0 + +roi_head.bbox_head.fc_reg.bias - torch.Size([320]): +NormalInit: mean=0, std=0.001, bias=0 + +roi_head.bbox_head.shared_fcs.0.weight - torch.Size([1024, 12544]): +XavierInit: gain=1, distribution=uniform, bias=0 + +roi_head.bbox_head.shared_fcs.0.bias - torch.Size([1024]): +XavierInit: gain=1, distribution=uniform, bias=0 + +roi_head.bbox_head.shared_fcs.1.weight - torch.Size([1024, 1024]): +XavierInit: gain=1, distribution=uniform, bias=0 + +roi_head.bbox_head.shared_fcs.1.bias - torch.Size([1024]): +XavierInit: gain=1, distribution=uniform, bias=0 + +roi_head.mask_head.convs.0.conv.weight - torch.Size([256, 256, 3, 3]): +Initialized by user-defined `init_weights` in ConvModule + +roi_head.mask_head.convs.0.conv.bias - torch.Size([256]): +The value is the same before and after calling `init_weights` of MaskRCNN + +roi_head.mask_head.convs.1.conv.weight - torch.Size([256, 256, 3, 3]): +Initialized by user-defined `init_weights` in ConvModule + +roi_head.mask_head.convs.1.conv.bias - torch.Size([256]): +The value is the same before and after calling `init_weights` of MaskRCNN + +roi_head.mask_head.convs.2.conv.weight - torch.Size([256, 256, 3, 3]): +Initialized by user-defined `init_weights` in ConvModule + +roi_head.mask_head.convs.2.conv.bias - torch.Size([256]): +The value is the same before and after calling `init_weights` of MaskRCNN + +roi_head.mask_head.convs.3.conv.weight - torch.Size([256, 256, 3, 3]): +Initialized by user-defined `init_weights` in ConvModule + +roi_head.mask_head.convs.3.conv.bias - torch.Size([256]): +The value is the same before and after calling `init_weights` of MaskRCNN + +roi_head.mask_head.upsample.weight - torch.Size([256, 256, 2, 2]): +Initialized by user-defined `init_weights` in FCNMaskHead + +roi_head.mask_head.upsample.bias - torch.Size([256]): +Initialized by user-defined `init_weights` in FCNMaskHead + +roi_head.mask_head.conv_logits.weight - torch.Size([80, 256, 1, 1]): +Initialized by user-defined `init_weights` in FCNMaskHead + +roi_head.mask_head.conv_logits.bias - torch.Size([80]): +Initialized by user-defined `init_weights` in FCNMaskHead +2024-05-31 12:16:04,810 - mmdet - INFO - {'num_layers': 24, 'layer_decay_rate': 0.85, 'skip_stride': [2, 2]} +2024-05-31 12:16:04,810 - mmdet - INFO - Build LayerDecayOptimizerConstructor 0.850000 - 26 +2024-05-31 12:16:04,822 - mmdet - INFO - Param groups = { + "layer_25_decay": { + "param_names": [ + "backbone.w1", + "backbone.w2", + "backbone.w3", + "backbone.interactions.0.interaction_units_12.branch2to1_proj.weight", + "backbone.interactions.0.interaction_units_12.branch2to1_injector.attn.sampling_offsets.weight", + "backbone.interactions.0.interaction_units_12.branch2to1_injector.attn.attention_weights.weight", + "backbone.interactions.0.interaction_units_12.branch2to1_injector.attn.value_proj.weight", + "backbone.interactions.0.interaction_units_12.branch2to1_injector.attn.output_proj.weight", + "backbone.interactions.0.interaction_units_12.branch2to1_injector.ffn.fc1.weight", + "backbone.interactions.0.interaction_units_12.branch2to1_injector.ffn.dwconv.dwconv.weight", + "backbone.interactions.0.interaction_units_12.branch2to1_injector.ffn.fc2.weight", + "backbone.interactions.0.interaction_units_12.branch1to2_proj.weight", + "backbone.interactions.0.interaction_units_12.branch1to2_injector.attn.sampling_offsets.weight", + "backbone.interactions.0.interaction_units_12.branch1to2_injector.attn.attention_weights.weight", + "backbone.interactions.0.interaction_units_12.branch1to2_injector.attn.value_proj.weight", + "backbone.interactions.0.interaction_units_12.branch1to2_injector.attn.output_proj.weight", + "backbone.interactions.0.interaction_units_12.branch1to2_injector.ffn.fc1.weight", + "backbone.interactions.0.interaction_units_12.branch1to2_injector.ffn.dwconv.dwconv.weight", + "backbone.interactions.0.interaction_units_12.branch1to2_injector.ffn.fc2.weight", + "backbone.interactions.0.interaction_units_23.branch2to1_proj.weight", + "backbone.interactions.0.interaction_units_23.branch2to1_injector.attn.sampling_offsets.weight", + "backbone.interactions.0.interaction_units_23.branch2to1_injector.attn.attention_weights.weight", + "backbone.interactions.0.interaction_units_23.branch2to1_injector.attn.value_proj.weight", + "backbone.interactions.0.interaction_units_23.branch2to1_injector.attn.output_proj.weight", + "backbone.interactions.0.interaction_units_23.branch2to1_injector.ffn.fc1.weight", + "backbone.interactions.0.interaction_units_23.branch2to1_injector.ffn.dwconv.dwconv.weight", + "backbone.interactions.0.interaction_units_23.branch2to1_injector.ffn.fc2.weight", + "backbone.interactions.0.interaction_units_23.branch1to2_proj.weight", + "backbone.interactions.0.interaction_units_23.branch1to2_injector.attn.sampling_offsets.weight", + "backbone.interactions.0.interaction_units_23.branch1to2_injector.attn.attention_weights.weight", + "backbone.interactions.0.interaction_units_23.branch1to2_injector.attn.value_proj.weight", + "backbone.interactions.0.interaction_units_23.branch1to2_injector.attn.output_proj.weight", + "backbone.interactions.0.interaction_units_23.branch1to2_injector.ffn.fc1.weight", + "backbone.interactions.0.interaction_units_23.branch1to2_injector.ffn.dwconv.dwconv.weight", + "backbone.interactions.0.interaction_units_23.branch1to2_injector.ffn.fc2.weight", + "backbone.interactions.1.interaction_units_12.branch2to1_proj.weight", + "backbone.interactions.1.interaction_units_12.branch2to1_injector.attn.sampling_offsets.weight", + "backbone.interactions.1.interaction_units_12.branch2to1_injector.attn.attention_weights.weight", + "backbone.interactions.1.interaction_units_12.branch2to1_injector.attn.value_proj.weight", + "backbone.interactions.1.interaction_units_12.branch2to1_injector.attn.output_proj.weight", + "backbone.interactions.1.interaction_units_12.branch2to1_injector.ffn.fc1.weight", + "backbone.interactions.1.interaction_units_12.branch2to1_injector.ffn.dwconv.dwconv.weight", + "backbone.interactions.1.interaction_units_12.branch2to1_injector.ffn.fc2.weight", + "backbone.interactions.1.interaction_units_12.branch1to2_proj.weight", + "backbone.interactions.1.interaction_units_12.branch1to2_injector.attn.sampling_offsets.weight", + "backbone.interactions.1.interaction_units_12.branch1to2_injector.attn.attention_weights.weight", + "backbone.interactions.1.interaction_units_12.branch1to2_injector.attn.value_proj.weight", + "backbone.interactions.1.interaction_units_12.branch1to2_injector.attn.output_proj.weight", + "backbone.interactions.1.interaction_units_12.branch1to2_injector.ffn.fc1.weight", + "backbone.interactions.1.interaction_units_12.branch1to2_injector.ffn.dwconv.dwconv.weight", + "backbone.interactions.1.interaction_units_12.branch1to2_injector.ffn.fc2.weight", + "backbone.interactions.1.interaction_units_23.branch2to1_proj.weight", + "backbone.interactions.1.interaction_units_23.branch2to1_injector.attn.sampling_offsets.weight", + "backbone.interactions.1.interaction_units_23.branch2to1_injector.attn.attention_weights.weight", + "backbone.interactions.1.interaction_units_23.branch2to1_injector.attn.value_proj.weight", + "backbone.interactions.1.interaction_units_23.branch2to1_injector.attn.output_proj.weight", + "backbone.interactions.1.interaction_units_23.branch2to1_injector.ffn.fc1.weight", + "backbone.interactions.1.interaction_units_23.branch2to1_injector.ffn.dwconv.dwconv.weight", + "backbone.interactions.1.interaction_units_23.branch2to1_injector.ffn.fc2.weight", + "backbone.interactions.1.interaction_units_23.branch1to2_proj.weight", + "backbone.interactions.1.interaction_units_23.branch1to2_injector.attn.sampling_offsets.weight", + "backbone.interactions.1.interaction_units_23.branch1to2_injector.attn.attention_weights.weight", + "backbone.interactions.1.interaction_units_23.branch1to2_injector.attn.value_proj.weight", + "backbone.interactions.1.interaction_units_23.branch1to2_injector.attn.output_proj.weight", + "backbone.interactions.1.interaction_units_23.branch1to2_injector.ffn.fc1.weight", + "backbone.interactions.1.interaction_units_23.branch1to2_injector.ffn.dwconv.dwconv.weight", + "backbone.interactions.1.interaction_units_23.branch1to2_injector.ffn.fc2.weight", + "backbone.interactions.2.interaction_units_12.branch2to1_proj.weight", + "backbone.interactions.2.interaction_units_12.branch2to1_injector.attn.sampling_offsets.weight", + "backbone.interactions.2.interaction_units_12.branch2to1_injector.attn.attention_weights.weight", + "backbone.interactions.2.interaction_units_12.branch2to1_injector.attn.value_proj.weight", + "backbone.interactions.2.interaction_units_12.branch2to1_injector.attn.output_proj.weight", + "backbone.interactions.2.interaction_units_12.branch2to1_injector.ffn.fc1.weight", + "backbone.interactions.2.interaction_units_12.branch2to1_injector.ffn.dwconv.dwconv.weight", + "backbone.interactions.2.interaction_units_12.branch2to1_injector.ffn.fc2.weight", + "backbone.interactions.2.interaction_units_12.branch1to2_proj.weight", + "backbone.interactions.2.interaction_units_12.branch1to2_injector.attn.sampling_offsets.weight", + "backbone.interactions.2.interaction_units_12.branch1to2_injector.attn.attention_weights.weight", + "backbone.interactions.2.interaction_units_12.branch1to2_injector.attn.value_proj.weight", + "backbone.interactions.2.interaction_units_12.branch1to2_injector.attn.output_proj.weight", + "backbone.interactions.2.interaction_units_12.branch1to2_injector.ffn.fc1.weight", + "backbone.interactions.2.interaction_units_12.branch1to2_injector.ffn.dwconv.dwconv.weight", + "backbone.interactions.2.interaction_units_12.branch1to2_injector.ffn.fc2.weight", + "backbone.interactions.2.interaction_units_23.branch2to1_proj.weight", + "backbone.interactions.2.interaction_units_23.branch2to1_injector.attn.sampling_offsets.weight", + "backbone.interactions.2.interaction_units_23.branch2to1_injector.attn.attention_weights.weight", + "backbone.interactions.2.interaction_units_23.branch2to1_injector.attn.value_proj.weight", + "backbone.interactions.2.interaction_units_23.branch2to1_injector.attn.output_proj.weight", + "backbone.interactions.2.interaction_units_23.branch2to1_injector.ffn.fc1.weight", + "backbone.interactions.2.interaction_units_23.branch2to1_injector.ffn.dwconv.dwconv.weight", + "backbone.interactions.2.interaction_units_23.branch2to1_injector.ffn.fc2.weight", + "backbone.interactions.2.interaction_units_23.branch1to2_proj.weight", + "backbone.interactions.2.interaction_units_23.branch1to2_injector.attn.sampling_offsets.weight", + "backbone.interactions.2.interaction_units_23.branch1to2_injector.attn.attention_weights.weight", + "backbone.interactions.2.interaction_units_23.branch1to2_injector.attn.value_proj.weight", + "backbone.interactions.2.interaction_units_23.branch1to2_injector.attn.output_proj.weight", + "backbone.interactions.2.interaction_units_23.branch1to2_injector.ffn.fc1.weight", + "backbone.interactions.2.interaction_units_23.branch1to2_injector.ffn.dwconv.dwconv.weight", + "backbone.interactions.2.interaction_units_23.branch1to2_injector.ffn.fc2.weight", + "backbone.interactions.3.interaction_units_12.branch2to1_proj.weight", + "backbone.interactions.3.interaction_units_12.branch2to1_injector.attn.sampling_offsets.weight", + "backbone.interactions.3.interaction_units_12.branch2to1_injector.attn.attention_weights.weight", + "backbone.interactions.3.interaction_units_12.branch2to1_injector.attn.value_proj.weight", + "backbone.interactions.3.interaction_units_12.branch2to1_injector.attn.output_proj.weight", + "backbone.interactions.3.interaction_units_12.branch2to1_injector.ffn.fc1.weight", + "backbone.interactions.3.interaction_units_12.branch2to1_injector.ffn.dwconv.dwconv.weight", + "backbone.interactions.3.interaction_units_12.branch2to1_injector.ffn.fc2.weight", + "backbone.interactions.3.interaction_units_12.branch1to2_proj.weight", + "backbone.interactions.3.interaction_units_12.branch1to2_injector.attn.sampling_offsets.weight", + "backbone.interactions.3.interaction_units_12.branch1to2_injector.attn.attention_weights.weight", + "backbone.interactions.3.interaction_units_12.branch1to2_injector.attn.value_proj.weight", + "backbone.interactions.3.interaction_units_12.branch1to2_injector.attn.output_proj.weight", + "backbone.interactions.3.interaction_units_12.branch1to2_injector.ffn.fc1.weight", + "backbone.interactions.3.interaction_units_12.branch1to2_injector.ffn.dwconv.dwconv.weight", + "backbone.interactions.3.interaction_units_12.branch1to2_injector.ffn.fc2.weight", + "backbone.interactions.3.interaction_units_23.branch2to1_proj.weight", + "backbone.interactions.3.interaction_units_23.branch2to1_injector.attn.sampling_offsets.weight", + "backbone.interactions.3.interaction_units_23.branch2to1_injector.attn.attention_weights.weight", + "backbone.interactions.3.interaction_units_23.branch2to1_injector.attn.value_proj.weight", + "backbone.interactions.3.interaction_units_23.branch2to1_injector.attn.output_proj.weight", + "backbone.interactions.3.interaction_units_23.branch2to1_injector.ffn.fc1.weight", + "backbone.interactions.3.interaction_units_23.branch2to1_injector.ffn.dwconv.dwconv.weight", + "backbone.interactions.3.interaction_units_23.branch2to1_injector.ffn.fc2.weight", + "backbone.interactions.3.interaction_units_23.branch1to2_proj.weight", + "backbone.interactions.3.interaction_units_23.branch1to2_injector.attn.sampling_offsets.weight", + "backbone.interactions.3.interaction_units_23.branch1to2_injector.attn.attention_weights.weight", + "backbone.interactions.3.interaction_units_23.branch1to2_injector.attn.value_proj.weight", + "backbone.interactions.3.interaction_units_23.branch1to2_injector.attn.output_proj.weight", + "backbone.interactions.3.interaction_units_23.branch1to2_injector.ffn.fc1.weight", + "backbone.interactions.3.interaction_units_23.branch1to2_injector.ffn.dwconv.dwconv.weight", + "backbone.interactions.3.interaction_units_23.branch1to2_injector.ffn.fc2.weight", + "backbone.interactions.4.interaction_units_12.branch2to1_proj.weight", + "backbone.interactions.4.interaction_units_12.branch2to1_injector.attn.sampling_offsets.weight", + "backbone.interactions.4.interaction_units_12.branch2to1_injector.attn.attention_weights.weight", + "backbone.interactions.4.interaction_units_12.branch2to1_injector.attn.value_proj.weight", + "backbone.interactions.4.interaction_units_12.branch2to1_injector.attn.output_proj.weight", + "backbone.interactions.4.interaction_units_12.branch2to1_injector.ffn.fc1.weight", + "backbone.interactions.4.interaction_units_12.branch2to1_injector.ffn.dwconv.dwconv.weight", + "backbone.interactions.4.interaction_units_12.branch2to1_injector.ffn.fc2.weight", + "backbone.interactions.4.interaction_units_12.branch1to2_proj.weight", + "backbone.interactions.4.interaction_units_12.branch1to2_injector.attn.sampling_offsets.weight", + "backbone.interactions.4.interaction_units_12.branch1to2_injector.attn.attention_weights.weight", + "backbone.interactions.4.interaction_units_12.branch1to2_injector.attn.value_proj.weight", + "backbone.interactions.4.interaction_units_12.branch1to2_injector.attn.output_proj.weight", + "backbone.interactions.4.interaction_units_12.branch1to2_injector.ffn.fc1.weight", + "backbone.interactions.4.interaction_units_12.branch1to2_injector.ffn.dwconv.dwconv.weight", + "backbone.interactions.4.interaction_units_12.branch1to2_injector.ffn.fc2.weight", + "backbone.interactions.4.interaction_units_23.branch2to1_proj.weight", + "backbone.interactions.4.interaction_units_23.branch2to1_injector.attn.sampling_offsets.weight", + "backbone.interactions.4.interaction_units_23.branch2to1_injector.attn.attention_weights.weight", + "backbone.interactions.4.interaction_units_23.branch2to1_injector.attn.value_proj.weight", + "backbone.interactions.4.interaction_units_23.branch2to1_injector.attn.output_proj.weight", + "backbone.interactions.4.interaction_units_23.branch2to1_injector.ffn.fc1.weight", + "backbone.interactions.4.interaction_units_23.branch2to1_injector.ffn.dwconv.dwconv.weight", + "backbone.interactions.4.interaction_units_23.branch2to1_injector.ffn.fc2.weight", + "backbone.interactions.4.interaction_units_23.branch1to2_proj.weight", + "backbone.interactions.4.interaction_units_23.branch1to2_injector.attn.sampling_offsets.weight", + "backbone.interactions.4.interaction_units_23.branch1to2_injector.attn.attention_weights.weight", + "backbone.interactions.4.interaction_units_23.branch1to2_injector.attn.value_proj.weight", + "backbone.interactions.4.interaction_units_23.branch1to2_injector.attn.output_proj.weight", + "backbone.interactions.4.interaction_units_23.branch1to2_injector.ffn.fc1.weight", + "backbone.interactions.4.interaction_units_23.branch1to2_injector.ffn.dwconv.dwconv.weight", + "backbone.interactions.4.interaction_units_23.branch1to2_injector.ffn.fc2.weight", + "backbone.interactions.5.interaction_units_12.branch2to1_proj.weight", + "backbone.interactions.5.interaction_units_12.branch2to1_injector.attn.sampling_offsets.weight", + "backbone.interactions.5.interaction_units_12.branch2to1_injector.attn.attention_weights.weight", + "backbone.interactions.5.interaction_units_12.branch2to1_injector.attn.value_proj.weight", + "backbone.interactions.5.interaction_units_12.branch2to1_injector.attn.output_proj.weight", + "backbone.interactions.5.interaction_units_12.branch2to1_injector.ffn.fc1.weight", + "backbone.interactions.5.interaction_units_12.branch2to1_injector.ffn.dwconv.dwconv.weight", + "backbone.interactions.5.interaction_units_12.branch2to1_injector.ffn.fc2.weight", + "backbone.interactions.5.interaction_units_12.branch1to2_proj.weight", + "backbone.interactions.5.interaction_units_12.branch1to2_injector.attn.sampling_offsets.weight", + "backbone.interactions.5.interaction_units_12.branch1to2_injector.attn.attention_weights.weight", + "backbone.interactions.5.interaction_units_12.branch1to2_injector.attn.value_proj.weight", + "backbone.interactions.5.interaction_units_12.branch1to2_injector.attn.output_proj.weight", + "backbone.interactions.5.interaction_units_12.branch1to2_injector.ffn.fc1.weight", + "backbone.interactions.5.interaction_units_12.branch1to2_injector.ffn.dwconv.dwconv.weight", + "backbone.interactions.5.interaction_units_12.branch1to2_injector.ffn.fc2.weight", + "backbone.interactions.5.interaction_units_23.branch2to1_proj.weight", + "backbone.interactions.5.interaction_units_23.branch2to1_injector.attn.sampling_offsets.weight", + "backbone.interactions.5.interaction_units_23.branch2to1_injector.attn.attention_weights.weight", + "backbone.interactions.5.interaction_units_23.branch2to1_injector.attn.value_proj.weight", + "backbone.interactions.5.interaction_units_23.branch2to1_injector.attn.output_proj.weight", + "backbone.interactions.5.interaction_units_23.branch2to1_injector.ffn.fc1.weight", + "backbone.interactions.5.interaction_units_23.branch2to1_injector.ffn.dwconv.dwconv.weight", + "backbone.interactions.5.interaction_units_23.branch2to1_injector.ffn.fc2.weight", + "backbone.interactions.5.interaction_units_23.branch1to2_proj.weight", + "backbone.interactions.5.interaction_units_23.branch1to2_injector.attn.sampling_offsets.weight", + "backbone.interactions.5.interaction_units_23.branch1to2_injector.attn.attention_weights.weight", + "backbone.interactions.5.interaction_units_23.branch1to2_injector.attn.value_proj.weight", + "backbone.interactions.5.interaction_units_23.branch1to2_injector.attn.output_proj.weight", + "backbone.interactions.5.interaction_units_23.branch1to2_injector.ffn.fc1.weight", + "backbone.interactions.5.interaction_units_23.branch1to2_injector.ffn.dwconv.dwconv.weight", + "backbone.interactions.5.interaction_units_23.branch1to2_injector.ffn.fc2.weight", + "backbone.interactions.6.interaction_units_12.branch2to1_proj.weight", + "backbone.interactions.6.interaction_units_12.branch2to1_injector.attn.sampling_offsets.weight", + "backbone.interactions.6.interaction_units_12.branch2to1_injector.attn.attention_weights.weight", + "backbone.interactions.6.interaction_units_12.branch2to1_injector.attn.value_proj.weight", + "backbone.interactions.6.interaction_units_12.branch2to1_injector.attn.output_proj.weight", + "backbone.interactions.6.interaction_units_12.branch2to1_injector.ffn.fc1.weight", + "backbone.interactions.6.interaction_units_12.branch2to1_injector.ffn.dwconv.dwconv.weight", + "backbone.interactions.6.interaction_units_12.branch2to1_injector.ffn.fc2.weight", + "backbone.interactions.6.interaction_units_12.branch1to2_proj.weight", + "backbone.interactions.6.interaction_units_12.branch1to2_injector.attn.sampling_offsets.weight", + "backbone.interactions.6.interaction_units_12.branch1to2_injector.attn.attention_weights.weight", + "backbone.interactions.6.interaction_units_12.branch1to2_injector.attn.value_proj.weight", + "backbone.interactions.6.interaction_units_12.branch1to2_injector.attn.output_proj.weight", + "backbone.interactions.6.interaction_units_12.branch1to2_injector.ffn.fc1.weight", + "backbone.interactions.6.interaction_units_12.branch1to2_injector.ffn.dwconv.dwconv.weight", + "backbone.interactions.6.interaction_units_12.branch1to2_injector.ffn.fc2.weight", + "backbone.interactions.6.interaction_units_23.branch2to1_proj.weight", + "backbone.interactions.6.interaction_units_23.branch2to1_injector.attn.sampling_offsets.weight", + "backbone.interactions.6.interaction_units_23.branch2to1_injector.attn.attention_weights.weight", + "backbone.interactions.6.interaction_units_23.branch2to1_injector.attn.value_proj.weight", + "backbone.interactions.6.interaction_units_23.branch2to1_injector.attn.output_proj.weight", + "backbone.interactions.6.interaction_units_23.branch2to1_injector.ffn.fc1.weight", + "backbone.interactions.6.interaction_units_23.branch2to1_injector.ffn.dwconv.dwconv.weight", + "backbone.interactions.6.interaction_units_23.branch2to1_injector.ffn.fc2.weight", + "backbone.interactions.6.interaction_units_23.branch1to2_proj.weight", + "backbone.interactions.6.interaction_units_23.branch1to2_injector.attn.sampling_offsets.weight", + "backbone.interactions.6.interaction_units_23.branch1to2_injector.attn.attention_weights.weight", + "backbone.interactions.6.interaction_units_23.branch1to2_injector.attn.value_proj.weight", + "backbone.interactions.6.interaction_units_23.branch1to2_injector.attn.output_proj.weight", + "backbone.interactions.6.interaction_units_23.branch1to2_injector.ffn.fc1.weight", + "backbone.interactions.6.interaction_units_23.branch1to2_injector.ffn.dwconv.dwconv.weight", + "backbone.interactions.6.interaction_units_23.branch1to2_injector.ffn.fc2.weight", + "backbone.interactions.7.interaction_units_12.branch2to1_proj.weight", + "backbone.interactions.7.interaction_units_12.branch2to1_injector.attn.sampling_offsets.weight", + "backbone.interactions.7.interaction_units_12.branch2to1_injector.attn.attention_weights.weight", + "backbone.interactions.7.interaction_units_12.branch2to1_injector.attn.value_proj.weight", + "backbone.interactions.7.interaction_units_12.branch2to1_injector.attn.output_proj.weight", + "backbone.interactions.7.interaction_units_12.branch2to1_injector.ffn.fc1.weight", + "backbone.interactions.7.interaction_units_12.branch2to1_injector.ffn.dwconv.dwconv.weight", + "backbone.interactions.7.interaction_units_12.branch2to1_injector.ffn.fc2.weight", + "backbone.interactions.7.interaction_units_12.branch1to2_proj.weight", + "backbone.interactions.7.interaction_units_12.branch1to2_injector.attn.sampling_offsets.weight", + "backbone.interactions.7.interaction_units_12.branch1to2_injector.attn.attention_weights.weight", + "backbone.interactions.7.interaction_units_12.branch1to2_injector.attn.value_proj.weight", + "backbone.interactions.7.interaction_units_12.branch1to2_injector.attn.output_proj.weight", + "backbone.interactions.7.interaction_units_12.branch1to2_injector.ffn.fc1.weight", + "backbone.interactions.7.interaction_units_12.branch1to2_injector.ffn.dwconv.dwconv.weight", + "backbone.interactions.7.interaction_units_12.branch1to2_injector.ffn.fc2.weight", + "backbone.interactions.7.interaction_units_23.branch2to1_proj.weight", + "backbone.interactions.7.interaction_units_23.branch2to1_injector.attn.sampling_offsets.weight", + "backbone.interactions.7.interaction_units_23.branch2to1_injector.attn.attention_weights.weight", + "backbone.interactions.7.interaction_units_23.branch2to1_injector.attn.value_proj.weight", + "backbone.interactions.7.interaction_units_23.branch2to1_injector.attn.output_proj.weight", + "backbone.interactions.7.interaction_units_23.branch2to1_injector.ffn.fc1.weight", + "backbone.interactions.7.interaction_units_23.branch2to1_injector.ffn.dwconv.dwconv.weight", + "backbone.interactions.7.interaction_units_23.branch2to1_injector.ffn.fc2.weight", + "backbone.interactions.7.interaction_units_23.branch1to2_proj.weight", + "backbone.interactions.7.interaction_units_23.branch1to2_injector.attn.sampling_offsets.weight", + "backbone.interactions.7.interaction_units_23.branch1to2_injector.attn.attention_weights.weight", + "backbone.interactions.7.interaction_units_23.branch1to2_injector.attn.value_proj.weight", + "backbone.interactions.7.interaction_units_23.branch1to2_injector.attn.output_proj.weight", + "backbone.interactions.7.interaction_units_23.branch1to2_injector.ffn.fc1.weight", + "backbone.interactions.7.interaction_units_23.branch1to2_injector.ffn.dwconv.dwconv.weight", + "backbone.interactions.7.interaction_units_23.branch1to2_injector.ffn.fc2.weight", + "backbone.interactions.8.interaction_units_12.branch2to1_proj.weight", + "backbone.interactions.8.interaction_units_12.branch2to1_injector.attn.sampling_offsets.weight", + "backbone.interactions.8.interaction_units_12.branch2to1_injector.attn.attention_weights.weight", + "backbone.interactions.8.interaction_units_12.branch2to1_injector.attn.value_proj.weight", + "backbone.interactions.8.interaction_units_12.branch2to1_injector.attn.output_proj.weight", + "backbone.interactions.8.interaction_units_12.branch2to1_injector.ffn.fc1.weight", + "backbone.interactions.8.interaction_units_12.branch2to1_injector.ffn.dwconv.dwconv.weight", + "backbone.interactions.8.interaction_units_12.branch2to1_injector.ffn.fc2.weight", + "backbone.interactions.8.interaction_units_12.branch1to2_proj.weight", + "backbone.interactions.8.interaction_units_12.branch1to2_injector.attn.sampling_offsets.weight", + "backbone.interactions.8.interaction_units_12.branch1to2_injector.attn.attention_weights.weight", + "backbone.interactions.8.interaction_units_12.branch1to2_injector.attn.value_proj.weight", + "backbone.interactions.8.interaction_units_12.branch1to2_injector.attn.output_proj.weight", + "backbone.interactions.8.interaction_units_12.branch1to2_injector.ffn.fc1.weight", + "backbone.interactions.8.interaction_units_12.branch1to2_injector.ffn.dwconv.dwconv.weight", + "backbone.interactions.8.interaction_units_12.branch1to2_injector.ffn.fc2.weight", + "backbone.interactions.8.interaction_units_23.branch2to1_proj.weight", + "backbone.interactions.8.interaction_units_23.branch2to1_injector.attn.sampling_offsets.weight", + "backbone.interactions.8.interaction_units_23.branch2to1_injector.attn.attention_weights.weight", + "backbone.interactions.8.interaction_units_23.branch2to1_injector.attn.value_proj.weight", + "backbone.interactions.8.interaction_units_23.branch2to1_injector.attn.output_proj.weight", + "backbone.interactions.8.interaction_units_23.branch2to1_injector.ffn.fc1.weight", + "backbone.interactions.8.interaction_units_23.branch2to1_injector.ffn.dwconv.dwconv.weight", + "backbone.interactions.8.interaction_units_23.branch2to1_injector.ffn.fc2.weight", + "backbone.interactions.8.interaction_units_23.branch1to2_proj.weight", + "backbone.interactions.8.interaction_units_23.branch1to2_injector.attn.sampling_offsets.weight", + "backbone.interactions.8.interaction_units_23.branch1to2_injector.attn.attention_weights.weight", + "backbone.interactions.8.interaction_units_23.branch1to2_injector.attn.value_proj.weight", + "backbone.interactions.8.interaction_units_23.branch1to2_injector.attn.output_proj.weight", + "backbone.interactions.8.interaction_units_23.branch1to2_injector.ffn.fc1.weight", + "backbone.interactions.8.interaction_units_23.branch1to2_injector.ffn.dwconv.dwconv.weight", + "backbone.interactions.8.interaction_units_23.branch1to2_injector.ffn.fc2.weight", + "backbone.interactions.9.interaction_units_12.branch2to1_proj.weight", + "backbone.interactions.9.interaction_units_12.branch2to1_injector.attn.sampling_offsets.weight", + "backbone.interactions.9.interaction_units_12.branch2to1_injector.attn.attention_weights.weight", + "backbone.interactions.9.interaction_units_12.branch2to1_injector.attn.value_proj.weight", + "backbone.interactions.9.interaction_units_12.branch2to1_injector.attn.output_proj.weight", + "backbone.interactions.9.interaction_units_12.branch2to1_injector.ffn.fc1.weight", + "backbone.interactions.9.interaction_units_12.branch2to1_injector.ffn.dwconv.dwconv.weight", + "backbone.interactions.9.interaction_units_12.branch2to1_injector.ffn.fc2.weight", + "backbone.interactions.9.interaction_units_12.branch1to2_proj.weight", + "backbone.interactions.9.interaction_units_12.branch1to2_injector.attn.sampling_offsets.weight", + "backbone.interactions.9.interaction_units_12.branch1to2_injector.attn.attention_weights.weight", + "backbone.interactions.9.interaction_units_12.branch1to2_injector.attn.value_proj.weight", + "backbone.interactions.9.interaction_units_12.branch1to2_injector.attn.output_proj.weight", + "backbone.interactions.9.interaction_units_12.branch1to2_injector.ffn.fc1.weight", + "backbone.interactions.9.interaction_units_12.branch1to2_injector.ffn.dwconv.dwconv.weight", + "backbone.interactions.9.interaction_units_12.branch1to2_injector.ffn.fc2.weight", + "backbone.interactions.9.interaction_units_23.branch2to1_proj.weight", + "backbone.interactions.9.interaction_units_23.branch2to1_injector.attn.sampling_offsets.weight", + "backbone.interactions.9.interaction_units_23.branch2to1_injector.attn.attention_weights.weight", + "backbone.interactions.9.interaction_units_23.branch2to1_injector.attn.value_proj.weight", + "backbone.interactions.9.interaction_units_23.branch2to1_injector.attn.output_proj.weight", + "backbone.interactions.9.interaction_units_23.branch2to1_injector.ffn.fc1.weight", + "backbone.interactions.9.interaction_units_23.branch2to1_injector.ffn.dwconv.dwconv.weight", + "backbone.interactions.9.interaction_units_23.branch2to1_injector.ffn.fc2.weight", + "backbone.interactions.9.interaction_units_23.branch1to2_proj.weight", + "backbone.interactions.9.interaction_units_23.branch1to2_injector.attn.sampling_offsets.weight", + "backbone.interactions.9.interaction_units_23.branch1to2_injector.attn.attention_weights.weight", + "backbone.interactions.9.interaction_units_23.branch1to2_injector.attn.value_proj.weight", + "backbone.interactions.9.interaction_units_23.branch1to2_injector.attn.output_proj.weight", + "backbone.interactions.9.interaction_units_23.branch1to2_injector.ffn.fc1.weight", + "backbone.interactions.9.interaction_units_23.branch1to2_injector.ffn.dwconv.dwconv.weight", + "backbone.interactions.9.interaction_units_23.branch1to2_injector.ffn.fc2.weight", + "backbone.interactions.10.interaction_units_12.branch2to1_proj.weight", + "backbone.interactions.10.interaction_units_12.branch2to1_injector.attn.sampling_offsets.weight", + "backbone.interactions.10.interaction_units_12.branch2to1_injector.attn.attention_weights.weight", + "backbone.interactions.10.interaction_units_12.branch2to1_injector.attn.value_proj.weight", + "backbone.interactions.10.interaction_units_12.branch2to1_injector.attn.output_proj.weight", + "backbone.interactions.10.interaction_units_12.branch2to1_injector.ffn.fc1.weight", + "backbone.interactions.10.interaction_units_12.branch2to1_injector.ffn.dwconv.dwconv.weight", + "backbone.interactions.10.interaction_units_12.branch2to1_injector.ffn.fc2.weight", + "backbone.interactions.10.interaction_units_12.branch1to2_proj.weight", + "backbone.interactions.10.interaction_units_12.branch1to2_injector.attn.sampling_offsets.weight", + "backbone.interactions.10.interaction_units_12.branch1to2_injector.attn.attention_weights.weight", + "backbone.interactions.10.interaction_units_12.branch1to2_injector.attn.value_proj.weight", + "backbone.interactions.10.interaction_units_12.branch1to2_injector.attn.output_proj.weight", + "backbone.interactions.10.interaction_units_12.branch1to2_injector.ffn.fc1.weight", + "backbone.interactions.10.interaction_units_12.branch1to2_injector.ffn.dwconv.dwconv.weight", + "backbone.interactions.10.interaction_units_12.branch1to2_injector.ffn.fc2.weight", + "backbone.interactions.10.interaction_units_23.branch2to1_proj.weight", + "backbone.interactions.10.interaction_units_23.branch2to1_injector.attn.sampling_offsets.weight", + "backbone.interactions.10.interaction_units_23.branch2to1_injector.attn.attention_weights.weight", + "backbone.interactions.10.interaction_units_23.branch2to1_injector.attn.value_proj.weight", + "backbone.interactions.10.interaction_units_23.branch2to1_injector.attn.output_proj.weight", + "backbone.interactions.10.interaction_units_23.branch2to1_injector.ffn.fc1.weight", + "backbone.interactions.10.interaction_units_23.branch2to1_injector.ffn.dwconv.dwconv.weight", + "backbone.interactions.10.interaction_units_23.branch2to1_injector.ffn.fc2.weight", + "backbone.interactions.10.interaction_units_23.branch1to2_proj.weight", + "backbone.interactions.10.interaction_units_23.branch1to2_injector.attn.sampling_offsets.weight", + "backbone.interactions.10.interaction_units_23.branch1to2_injector.attn.attention_weights.weight", + "backbone.interactions.10.interaction_units_23.branch1to2_injector.attn.value_proj.weight", + "backbone.interactions.10.interaction_units_23.branch1to2_injector.attn.output_proj.weight", + "backbone.interactions.10.interaction_units_23.branch1to2_injector.ffn.fc1.weight", + "backbone.interactions.10.interaction_units_23.branch1to2_injector.ffn.dwconv.dwconv.weight", + "backbone.interactions.10.interaction_units_23.branch1to2_injector.ffn.fc2.weight", + "backbone.interactions.11.interaction_units_12.branch2to1_proj.weight", + "backbone.interactions.11.interaction_units_12.branch2to1_injector.attn.sampling_offsets.weight", + "backbone.interactions.11.interaction_units_12.branch2to1_injector.attn.attention_weights.weight", + "backbone.interactions.11.interaction_units_12.branch2to1_injector.attn.value_proj.weight", + "backbone.interactions.11.interaction_units_12.branch2to1_injector.attn.output_proj.weight", + "backbone.interactions.11.interaction_units_12.branch2to1_injector.ffn.fc1.weight", + "backbone.interactions.11.interaction_units_12.branch2to1_injector.ffn.dwconv.dwconv.weight", + "backbone.interactions.11.interaction_units_12.branch2to1_injector.ffn.fc2.weight", + "backbone.interactions.11.interaction_units_12.branch1to2_proj.weight", + "backbone.interactions.11.interaction_units_12.branch1to2_injector.attn.sampling_offsets.weight", + "backbone.interactions.11.interaction_units_12.branch1to2_injector.attn.attention_weights.weight", + "backbone.interactions.11.interaction_units_12.branch1to2_injector.attn.value_proj.weight", + "backbone.interactions.11.interaction_units_12.branch1to2_injector.attn.output_proj.weight", + "backbone.interactions.11.interaction_units_12.branch1to2_injector.ffn.fc1.weight", + "backbone.interactions.11.interaction_units_12.branch1to2_injector.ffn.dwconv.dwconv.weight", + "backbone.interactions.11.interaction_units_12.branch1to2_injector.ffn.fc2.weight", + "backbone.interactions.11.interaction_units_23.branch2to1_proj.weight", + "backbone.interactions.11.interaction_units_23.branch2to1_injector.attn.sampling_offsets.weight", + "backbone.interactions.11.interaction_units_23.branch2to1_injector.attn.attention_weights.weight", + "backbone.interactions.11.interaction_units_23.branch2to1_injector.attn.value_proj.weight", + "backbone.interactions.11.interaction_units_23.branch2to1_injector.attn.output_proj.weight", + "backbone.interactions.11.interaction_units_23.branch2to1_injector.ffn.fc1.weight", + "backbone.interactions.11.interaction_units_23.branch2to1_injector.ffn.dwconv.dwconv.weight", + "backbone.interactions.11.interaction_units_23.branch2to1_injector.ffn.fc2.weight", + "backbone.interactions.11.interaction_units_23.branch1to2_proj.weight", + "backbone.interactions.11.interaction_units_23.branch1to2_injector.attn.sampling_offsets.weight", + "backbone.interactions.11.interaction_units_23.branch1to2_injector.attn.attention_weights.weight", + "backbone.interactions.11.interaction_units_23.branch1to2_injector.attn.value_proj.weight", + "backbone.interactions.11.interaction_units_23.branch1to2_injector.attn.output_proj.weight", + "backbone.interactions.11.interaction_units_23.branch1to2_injector.ffn.fc1.weight", + "backbone.interactions.11.interaction_units_23.branch1to2_injector.ffn.dwconv.dwconv.weight", + "backbone.interactions.11.interaction_units_23.branch1to2_injector.ffn.fc2.weight", + "backbone.merge_branch1.0.weight", + "backbone.merge_branch1.3.weight", + "backbone.merge_branch2.0.weight", + "backbone.merge_branch2.3.weight", + "backbone.merge_branch3.0.weight", + "backbone.merge_branch3.3.weight", + "backbone.fpn1.0.weight", + "backbone.fpn1.3.weight", + "backbone.fpn2.0.weight", + "neck.lateral_convs.0.conv.weight", + "neck.lateral_convs.1.conv.weight", + "neck.lateral_convs.2.conv.weight", + "neck.lateral_convs.3.conv.weight", + "neck.fpn_convs.0.conv.weight", + "neck.fpn_convs.1.conv.weight", + "neck.fpn_convs.2.conv.weight", + "neck.fpn_convs.3.conv.weight", + "rpn_head.rpn_conv.weight", + "rpn_head.rpn_cls.weight", + "rpn_head.rpn_reg.weight", + "roi_head.bbox_head.fc_cls.weight", + "roi_head.bbox_head.fc_reg.weight", + "roi_head.bbox_head.shared_fcs.0.weight", + "roi_head.bbox_head.shared_fcs.1.weight", + "roi_head.mask_head.convs.0.conv.weight", + "roi_head.mask_head.convs.1.conv.weight", + "roi_head.mask_head.convs.2.conv.weight", + "roi_head.mask_head.convs.3.conv.weight", + "roi_head.mask_head.upsample.weight", + "roi_head.mask_head.conv_logits.weight" + ], + "lr_scale": 1.0, + "lr": 0.0001, + "weight_decay": 0.05 + }, + "layer_0_decay": { + "param_names": [ + "backbone.branch1.patch_embed.proj.weight", + "backbone.branch2.patch_embed.proj.weight", + "backbone.branch3.pos_embed", + "backbone.branch3.patch_embed.proj.weight" + ], + "lr_scale": 0.017197809852207896, + "lr": 1.7197809852207897e-06, + "weight_decay": 0.05 + }, + "layer_0_no_decay": { + "param_names": [ + "backbone.branch1.patch_embed.proj.bias", + "backbone.branch2.patch_embed.proj.bias", + "backbone.branch3.patch_embed.proj.bias" + ], + "lr_scale": 0.017197809852207896, + "lr": 1.7197809852207897e-06, + "weight_decay": 0.0 + }, + "layer_1_no_decay": { + "param_names": [ + "backbone.branch1.blocks.0.gamma_1", + "backbone.branch1.blocks.0.gamma_2", + "backbone.branch1.blocks.0.norm1.weight", + "backbone.branch1.blocks.0.norm1.bias", + "backbone.branch1.blocks.0.attn.q_bias", + "backbone.branch1.blocks.0.attn.v_bias", + "backbone.branch1.blocks.0.attn.proj.bias", + "backbone.branch1.blocks.0.norm2.weight", + "backbone.branch1.blocks.0.norm2.bias", + "backbone.branch1.blocks.0.mlp.fc1.bias", + "backbone.branch1.blocks.0.mlp.fc2.bias" + ], + "lr_scale": 0.02023271747318576, + "lr": 2.023271747318576e-06, + "weight_decay": 0.0 + }, + "layer_1_decay": { + "param_names": [ + "backbone.branch1.blocks.0.attn.relative_position_bias_table", + "backbone.branch1.blocks.0.attn.qkv.weight", + "backbone.branch1.blocks.0.attn.proj.weight", + "backbone.branch1.blocks.0.mlp.fc1.weight", + "backbone.branch1.blocks.0.mlp.fc2.weight" + ], + "lr_scale": 0.02023271747318576, + "lr": 2.023271747318576e-06, + "weight_decay": 0.05 + }, + "layer_2_no_decay": { + "param_names": [ + "backbone.branch1.blocks.1.gamma_1", + "backbone.branch1.blocks.1.gamma_2", + "backbone.branch1.blocks.1.norm1.weight", + "backbone.branch1.blocks.1.norm1.bias", + "backbone.branch1.blocks.1.attn.q_bias", + "backbone.branch1.blocks.1.attn.v_bias", + "backbone.branch1.blocks.1.attn.proj.bias", + "backbone.branch1.blocks.1.norm2.weight", + "backbone.branch1.blocks.1.norm2.bias", + "backbone.branch1.blocks.1.mlp.fc1.bias", + "backbone.branch1.blocks.1.mlp.fc2.bias", + "backbone.branch2.blocks.0.gamma_1", + "backbone.branch2.blocks.0.gamma_2", + "backbone.branch2.blocks.0.norm1.weight", + "backbone.branch2.blocks.0.norm1.bias", + "backbone.branch2.blocks.0.attn.q_bias", + "backbone.branch2.blocks.0.attn.v_bias", + "backbone.branch2.blocks.0.attn.proj.bias", + "backbone.branch2.blocks.0.norm2.weight", + "backbone.branch2.blocks.0.norm2.bias", + "backbone.branch2.blocks.0.mlp.fc1.bias", + "backbone.branch2.blocks.0.mlp.fc2.bias", + "backbone.branch3.blocks.0.gamma_1", + "backbone.branch3.blocks.0.gamma_2", + "backbone.branch3.blocks.0.norm1.weight", + "backbone.branch3.blocks.0.norm1.bias", + "backbone.branch3.blocks.0.attn.qkv.bias", + "backbone.branch3.blocks.0.attn.proj.bias", + "backbone.branch3.blocks.0.norm2.weight", + "backbone.branch3.blocks.0.norm2.bias", + "backbone.branch3.blocks.0.mlp.fc1.bias", + "backbone.branch3.blocks.0.mlp.fc2.bias" + ], + "lr_scale": 0.023803197027277366, + "lr": 2.380319702727737e-06, + "weight_decay": 0.0 + }, + "layer_2_decay": { + "param_names": [ + "backbone.branch1.blocks.1.attn.relative_position_bias_table", + "backbone.branch1.blocks.1.attn.qkv.weight", + "backbone.branch1.blocks.1.attn.proj.weight", + "backbone.branch1.blocks.1.mlp.fc1.weight", + "backbone.branch1.blocks.1.mlp.fc2.weight", + "backbone.branch2.blocks.0.attn.relative_position_bias_table", + "backbone.branch2.blocks.0.attn.qkv.weight", + "backbone.branch2.blocks.0.attn.proj.weight", + "backbone.branch2.blocks.0.mlp.fc1.weight", + "backbone.branch2.blocks.0.mlp.fc2.weight", + "backbone.branch3.blocks.0.attn.qkv.weight", + "backbone.branch3.blocks.0.attn.proj.weight", + "backbone.branch3.blocks.0.mlp.fc1.weight", + "backbone.branch3.blocks.0.mlp.fc2.weight" + ], + "lr_scale": 0.023803197027277366, + "lr": 2.380319702727737e-06, + "weight_decay": 0.05 + }, + "layer_3_no_decay": { + "param_names": [ + "backbone.branch1.blocks.2.gamma_1", + "backbone.branch1.blocks.2.gamma_2", + "backbone.branch1.blocks.2.norm1.weight", + "backbone.branch1.blocks.2.norm1.bias", + "backbone.branch1.blocks.2.attn.q_bias", + "backbone.branch1.blocks.2.attn.v_bias", + "backbone.branch1.blocks.2.attn.proj.bias", + "backbone.branch1.blocks.2.norm2.weight", + "backbone.branch1.blocks.2.norm2.bias", + "backbone.branch1.blocks.2.mlp.fc1.bias", + "backbone.branch1.blocks.2.mlp.fc2.bias" + ], + "lr_scale": 0.028003761208561607, + "lr": 2.8003761208561607e-06, + "weight_decay": 0.0 + }, + "layer_3_decay": { + "param_names": [ + "backbone.branch1.blocks.2.attn.relative_position_bias_table", + "backbone.branch1.blocks.2.attn.qkv.weight", + "backbone.branch1.blocks.2.attn.proj.weight", + "backbone.branch1.blocks.2.mlp.fc1.weight", + "backbone.branch1.blocks.2.mlp.fc2.weight" + ], + "lr_scale": 0.028003761208561607, + "lr": 2.8003761208561607e-06, + "weight_decay": 0.05 + }, + "layer_4_no_decay": { + "param_names": [ + "backbone.branch1.blocks.3.gamma_1", + "backbone.branch1.blocks.3.gamma_2", + "backbone.branch1.blocks.3.norm1.weight", + "backbone.branch1.blocks.3.norm1.bias", + "backbone.branch1.blocks.3.attn.q_bias", + "backbone.branch1.blocks.3.attn.v_bias", + "backbone.branch1.blocks.3.attn.proj.bias", + "backbone.branch1.blocks.3.norm2.weight", + "backbone.branch1.blocks.3.norm2.bias", + "backbone.branch1.blocks.3.mlp.fc1.bias", + "backbone.branch1.blocks.3.mlp.fc2.bias", + "backbone.branch2.blocks.1.gamma_1", + "backbone.branch2.blocks.1.gamma_2", + "backbone.branch2.blocks.1.norm1.weight", + "backbone.branch2.blocks.1.norm1.bias", + "backbone.branch2.blocks.1.attn.q_bias", + "backbone.branch2.blocks.1.attn.v_bias", + "backbone.branch2.blocks.1.attn.proj.bias", + "backbone.branch2.blocks.1.norm2.weight", + "backbone.branch2.blocks.1.norm2.bias", + "backbone.branch2.blocks.1.mlp.fc1.bias", + "backbone.branch2.blocks.1.mlp.fc2.bias", + "backbone.branch3.blocks.1.gamma_1", + "backbone.branch3.blocks.1.gamma_2", + "backbone.branch3.blocks.1.norm1.weight", + "backbone.branch3.blocks.1.norm1.bias", + "backbone.branch3.blocks.1.attn.qkv.bias", + "backbone.branch3.blocks.1.attn.proj.bias", + "backbone.branch3.blocks.1.norm2.weight", + "backbone.branch3.blocks.1.norm2.bias", + "backbone.branch3.blocks.1.mlp.fc1.bias", + "backbone.branch3.blocks.1.mlp.fc2.bias" + ], + "lr_scale": 0.03294560142183718, + "lr": 3.2945601421837183e-06, + "weight_decay": 0.0 + }, + "layer_4_decay": { + "param_names": [ + "backbone.branch1.blocks.3.attn.relative_position_bias_table", + "backbone.branch1.blocks.3.attn.qkv.weight", + "backbone.branch1.blocks.3.attn.proj.weight", + "backbone.branch1.blocks.3.mlp.fc1.weight", + "backbone.branch1.blocks.3.mlp.fc2.weight", + "backbone.branch2.blocks.1.attn.relative_position_bias_table", + "backbone.branch2.blocks.1.attn.qkv.weight", + "backbone.branch2.blocks.1.attn.proj.weight", + "backbone.branch2.blocks.1.mlp.fc1.weight", + "backbone.branch2.blocks.1.mlp.fc2.weight", + "backbone.branch3.blocks.1.attn.qkv.weight", + "backbone.branch3.blocks.1.attn.proj.weight", + "backbone.branch3.blocks.1.mlp.fc1.weight", + "backbone.branch3.blocks.1.mlp.fc2.weight" + ], + "lr_scale": 0.03294560142183718, + "lr": 3.2945601421837183e-06, + "weight_decay": 0.05 + }, + "layer_5_no_decay": { + "param_names": [ + "backbone.branch1.blocks.4.gamma_1", + "backbone.branch1.blocks.4.gamma_2", + "backbone.branch1.blocks.4.norm1.weight", + "backbone.branch1.blocks.4.norm1.bias", + "backbone.branch1.blocks.4.attn.q_bias", + "backbone.branch1.blocks.4.attn.v_bias", + "backbone.branch1.blocks.4.attn.proj.bias", + "backbone.branch1.blocks.4.norm2.weight", + "backbone.branch1.blocks.4.norm2.bias", + "backbone.branch1.blocks.4.mlp.fc1.bias", + "backbone.branch1.blocks.4.mlp.fc2.bias" + ], + "lr_scale": 0.03875953108451433, + "lr": 3.875953108451433e-06, + "weight_decay": 0.0 + }, + "layer_5_decay": { + "param_names": [ + "backbone.branch1.blocks.4.attn.relative_position_bias_table", + "backbone.branch1.blocks.4.attn.qkv.weight", + "backbone.branch1.blocks.4.attn.proj.weight", + "backbone.branch1.blocks.4.mlp.fc1.weight", + "backbone.branch1.blocks.4.mlp.fc2.weight" + ], + "lr_scale": 0.03875953108451433, + "lr": 3.875953108451433e-06, + "weight_decay": 0.05 + }, + "layer_6_no_decay": { + "param_names": [ + "backbone.branch1.blocks.5.gamma_1", + "backbone.branch1.blocks.5.gamma_2", + "backbone.branch1.blocks.5.norm1.weight", + "backbone.branch1.blocks.5.norm1.bias", + "backbone.branch1.blocks.5.attn.q_bias", + "backbone.branch1.blocks.5.attn.v_bias", + "backbone.branch1.blocks.5.attn.proj.bias", + "backbone.branch1.blocks.5.norm2.weight", + "backbone.branch1.blocks.5.norm2.bias", + "backbone.branch1.blocks.5.mlp.fc1.bias", + "backbone.branch1.blocks.5.mlp.fc2.bias", + "backbone.branch2.blocks.2.gamma_1", + "backbone.branch2.blocks.2.gamma_2", + "backbone.branch2.blocks.2.norm1.weight", + "backbone.branch2.blocks.2.norm1.bias", + "backbone.branch2.blocks.2.attn.q_bias", + "backbone.branch2.blocks.2.attn.v_bias", + "backbone.branch2.blocks.2.attn.proj.bias", + "backbone.branch2.blocks.2.norm2.weight", + "backbone.branch2.blocks.2.norm2.bias", + "backbone.branch2.blocks.2.mlp.fc1.bias", + "backbone.branch2.blocks.2.mlp.fc2.bias", + "backbone.branch3.blocks.2.gamma_1", + "backbone.branch3.blocks.2.gamma_2", + "backbone.branch3.blocks.2.norm1.weight", + "backbone.branch3.blocks.2.norm1.bias", + "backbone.branch3.blocks.2.attn.qkv.bias", + "backbone.branch3.blocks.2.attn.proj.bias", + "backbone.branch3.blocks.2.norm2.weight", + "backbone.branch3.blocks.2.norm2.bias", + "backbone.branch3.blocks.2.mlp.fc1.bias", + "backbone.branch3.blocks.2.mlp.fc2.bias" + ], + "lr_scale": 0.04559944833472275, + "lr": 4.5599448334722756e-06, + "weight_decay": 0.0 + }, + "layer_6_decay": { + "param_names": [ + "backbone.branch1.blocks.5.attn.relative_position_bias_table", + "backbone.branch1.blocks.5.attn.qkv.weight", + "backbone.branch1.blocks.5.attn.proj.weight", + "backbone.branch1.blocks.5.mlp.fc1.weight", + "backbone.branch1.blocks.5.mlp.fc2.weight", + "backbone.branch2.blocks.2.attn.relative_position_bias_table", + "backbone.branch2.blocks.2.attn.qkv.weight", + "backbone.branch2.blocks.2.attn.proj.weight", + "backbone.branch2.blocks.2.mlp.fc1.weight", + "backbone.branch2.blocks.2.mlp.fc2.weight", + "backbone.branch3.blocks.2.attn.qkv.weight", + "backbone.branch3.blocks.2.attn.proj.weight", + "backbone.branch3.blocks.2.mlp.fc1.weight", + "backbone.branch3.blocks.2.mlp.fc2.weight" + ], + "lr_scale": 0.04559944833472275, + "lr": 4.5599448334722756e-06, + "weight_decay": 0.05 + }, + "layer_7_no_decay": { + "param_names": [ + "backbone.branch1.blocks.6.gamma_1", + "backbone.branch1.blocks.6.gamma_2", + "backbone.branch1.blocks.6.norm1.weight", + "backbone.branch1.blocks.6.norm1.bias", + "backbone.branch1.blocks.6.attn.q_bias", + "backbone.branch1.blocks.6.attn.v_bias", + "backbone.branch1.blocks.6.attn.proj.bias", + "backbone.branch1.blocks.6.norm2.weight", + "backbone.branch1.blocks.6.norm2.bias", + "backbone.branch1.blocks.6.mlp.fc1.bias", + "backbone.branch1.blocks.6.mlp.fc2.bias" + ], + "lr_scale": 0.053646409805556176, + "lr": 5.364640980555618e-06, + "weight_decay": 0.0 + }, + "layer_7_decay": { + "param_names": [ + "backbone.branch1.blocks.6.attn.relative_position_bias_table", + "backbone.branch1.blocks.6.attn.qkv.weight", + "backbone.branch1.blocks.6.attn.proj.weight", + "backbone.branch1.blocks.6.mlp.fc1.weight", + "backbone.branch1.blocks.6.mlp.fc2.weight" + ], + "lr_scale": 0.053646409805556176, + "lr": 5.364640980555618e-06, + "weight_decay": 0.05 + }, + "layer_8_no_decay": { + "param_names": [ + "backbone.branch1.blocks.7.gamma_1", + "backbone.branch1.blocks.7.gamma_2", + "backbone.branch1.blocks.7.norm1.weight", + "backbone.branch1.blocks.7.norm1.bias", + "backbone.branch1.blocks.7.attn.q_bias", + "backbone.branch1.blocks.7.attn.v_bias", + "backbone.branch1.blocks.7.attn.proj.bias", + "backbone.branch1.blocks.7.norm2.weight", + "backbone.branch1.blocks.7.norm2.bias", + "backbone.branch1.blocks.7.mlp.fc1.bias", + "backbone.branch1.blocks.7.mlp.fc2.bias", + "backbone.branch2.blocks.3.gamma_1", + "backbone.branch2.blocks.3.gamma_2", + "backbone.branch2.blocks.3.norm1.weight", + "backbone.branch2.blocks.3.norm1.bias", + "backbone.branch2.blocks.3.attn.q_bias", + "backbone.branch2.blocks.3.attn.v_bias", + "backbone.branch2.blocks.3.attn.proj.bias", + "backbone.branch2.blocks.3.norm2.weight", + "backbone.branch2.blocks.3.norm2.bias", + "backbone.branch2.blocks.3.mlp.fc1.bias", + "backbone.branch2.blocks.3.mlp.fc2.bias", + "backbone.branch3.blocks.3.gamma_1", + "backbone.branch3.blocks.3.gamma_2", + "backbone.branch3.blocks.3.norm1.weight", + "backbone.branch3.blocks.3.norm1.bias", + "backbone.branch3.blocks.3.attn.qkv.bias", + "backbone.branch3.blocks.3.attn.proj.bias", + "backbone.branch3.blocks.3.norm2.weight", + "backbone.branch3.blocks.3.norm2.bias", + "backbone.branch3.blocks.3.mlp.fc1.bias", + "backbone.branch3.blocks.3.mlp.fc2.bias" + ], + "lr_scale": 0.06311342330065432, + "lr": 6.3113423300654325e-06, + "weight_decay": 0.0 + }, + "layer_8_decay": { + "param_names": [ + "backbone.branch1.blocks.7.attn.relative_position_bias_table", + "backbone.branch1.blocks.7.attn.qkv.weight", + "backbone.branch1.blocks.7.attn.proj.weight", + "backbone.branch1.blocks.7.mlp.fc1.weight", + "backbone.branch1.blocks.7.mlp.fc2.weight", + "backbone.branch2.blocks.3.attn.relative_position_bias_table", + "backbone.branch2.blocks.3.attn.qkv.weight", + "backbone.branch2.blocks.3.attn.proj.weight", + "backbone.branch2.blocks.3.mlp.fc1.weight", + "backbone.branch2.blocks.3.mlp.fc2.weight", + "backbone.branch3.blocks.3.attn.qkv.weight", + "backbone.branch3.blocks.3.attn.proj.weight", + "backbone.branch3.blocks.3.mlp.fc1.weight", + "backbone.branch3.blocks.3.mlp.fc2.weight" + ], + "lr_scale": 0.06311342330065432, + "lr": 6.3113423300654325e-06, + "weight_decay": 0.05 + }, + "layer_9_no_decay": { + "param_names": [ + "backbone.branch1.blocks.8.gamma_1", + "backbone.branch1.blocks.8.gamma_2", + "backbone.branch1.blocks.8.norm1.weight", + "backbone.branch1.blocks.8.norm1.bias", + "backbone.branch1.blocks.8.attn.q_bias", + "backbone.branch1.blocks.8.attn.v_bias", + "backbone.branch1.blocks.8.attn.proj.bias", + "backbone.branch1.blocks.8.norm2.weight", + "backbone.branch1.blocks.8.norm2.bias", + "backbone.branch1.blocks.8.mlp.fc1.bias", + "backbone.branch1.blocks.8.mlp.fc2.bias" + ], + "lr_scale": 0.07425108623606391, + "lr": 7.425108623606392e-06, + "weight_decay": 0.0 + }, + "layer_9_decay": { + "param_names": [ + "backbone.branch1.blocks.8.attn.relative_position_bias_table", + "backbone.branch1.blocks.8.attn.qkv.weight", + "backbone.branch1.blocks.8.attn.proj.weight", + "backbone.branch1.blocks.8.mlp.fc1.weight", + "backbone.branch1.blocks.8.mlp.fc2.weight" + ], + "lr_scale": 0.07425108623606391, + "lr": 7.425108623606392e-06, + "weight_decay": 0.05 + }, + "layer_10_no_decay": { + "param_names": [ + "backbone.branch1.blocks.9.gamma_1", + "backbone.branch1.blocks.9.gamma_2", + "backbone.branch1.blocks.9.norm1.weight", + "backbone.branch1.blocks.9.norm1.bias", + "backbone.branch1.blocks.9.attn.q_bias", + "backbone.branch1.blocks.9.attn.v_bias", + "backbone.branch1.blocks.9.attn.proj.bias", + "backbone.branch1.blocks.9.norm2.weight", + "backbone.branch1.blocks.9.norm2.bias", + "backbone.branch1.blocks.9.mlp.fc1.bias", + "backbone.branch1.blocks.9.mlp.fc2.bias", + "backbone.branch2.blocks.4.gamma_1", + "backbone.branch2.blocks.4.gamma_2", + "backbone.branch2.blocks.4.norm1.weight", + "backbone.branch2.blocks.4.norm1.bias", + "backbone.branch2.blocks.4.attn.q_bias", + "backbone.branch2.blocks.4.attn.v_bias", + "backbone.branch2.blocks.4.attn.proj.bias", + "backbone.branch2.blocks.4.norm2.weight", + "backbone.branch2.blocks.4.norm2.bias", + "backbone.branch2.blocks.4.mlp.fc1.bias", + "backbone.branch2.blocks.4.mlp.fc2.bias", + "backbone.branch3.blocks.4.gamma_1", + "backbone.branch3.blocks.4.gamma_2", + "backbone.branch3.blocks.4.norm1.weight", + "backbone.branch3.blocks.4.norm1.bias", + "backbone.branch3.blocks.4.attn.qkv.bias", + "backbone.branch3.blocks.4.attn.proj.bias", + "backbone.branch3.blocks.4.norm2.weight", + "backbone.branch3.blocks.4.norm2.bias", + "backbone.branch3.blocks.4.mlp.fc1.bias", + "backbone.branch3.blocks.4.mlp.fc2.bias" + ], + "lr_scale": 0.08735421910125167, + "lr": 8.735421910125167e-06, + "weight_decay": 0.0 + }, + "layer_10_decay": { + "param_names": [ + "backbone.branch1.blocks.9.attn.relative_position_bias_table", + "backbone.branch1.blocks.9.attn.qkv.weight", + "backbone.branch1.blocks.9.attn.proj.weight", + "backbone.branch1.blocks.9.mlp.fc1.weight", + "backbone.branch1.blocks.9.mlp.fc2.weight", + "backbone.branch2.blocks.4.attn.relative_position_bias_table", + "backbone.branch2.blocks.4.attn.qkv.weight", + "backbone.branch2.blocks.4.attn.proj.weight", + "backbone.branch2.blocks.4.mlp.fc1.weight", + "backbone.branch2.blocks.4.mlp.fc2.weight", + "backbone.branch3.blocks.4.attn.qkv.weight", + "backbone.branch3.blocks.4.attn.proj.weight", + "backbone.branch3.blocks.4.mlp.fc1.weight", + "backbone.branch3.blocks.4.mlp.fc2.weight" + ], + "lr_scale": 0.08735421910125167, + "lr": 8.735421910125167e-06, + "weight_decay": 0.05 + }, + "layer_11_no_decay": { + "param_names": [ + "backbone.branch1.blocks.10.gamma_1", + "backbone.branch1.blocks.10.gamma_2", + "backbone.branch1.blocks.10.norm1.weight", + "backbone.branch1.blocks.10.norm1.bias", + "backbone.branch1.blocks.10.attn.q_bias", + "backbone.branch1.blocks.10.attn.v_bias", + "backbone.branch1.blocks.10.attn.proj.bias", + "backbone.branch1.blocks.10.norm2.weight", + "backbone.branch1.blocks.10.norm2.bias", + "backbone.branch1.blocks.10.mlp.fc1.bias", + "backbone.branch1.blocks.10.mlp.fc2.bias" + ], + "lr_scale": 0.10276966953088432, + "lr": 1.0276966953088432e-05, + "weight_decay": 0.0 + }, + "layer_11_decay": { + "param_names": [ + "backbone.branch1.blocks.10.attn.relative_position_bias_table", + "backbone.branch1.blocks.10.attn.qkv.weight", + "backbone.branch1.blocks.10.attn.proj.weight", + "backbone.branch1.blocks.10.mlp.fc1.weight", + "backbone.branch1.blocks.10.mlp.fc2.weight" + ], + "lr_scale": 0.10276966953088432, + "lr": 1.0276966953088432e-05, + "weight_decay": 0.05 + }, + "layer_12_no_decay": { + "param_names": [ + "backbone.branch1.blocks.11.gamma_1", + "backbone.branch1.blocks.11.gamma_2", + "backbone.branch1.blocks.11.norm1.weight", + "backbone.branch1.blocks.11.norm1.bias", + "backbone.branch1.blocks.11.attn.q_bias", + "backbone.branch1.blocks.11.attn.v_bias", + "backbone.branch1.blocks.11.attn.proj.bias", + "backbone.branch1.blocks.11.norm2.weight", + "backbone.branch1.blocks.11.norm2.bias", + "backbone.branch1.blocks.11.mlp.fc1.bias", + "backbone.branch1.blocks.11.mlp.fc2.bias", + "backbone.branch2.blocks.5.gamma_1", + "backbone.branch2.blocks.5.gamma_2", + "backbone.branch2.blocks.5.norm1.weight", + "backbone.branch2.blocks.5.norm1.bias", + "backbone.branch2.blocks.5.attn.q_bias", + "backbone.branch2.blocks.5.attn.v_bias", + "backbone.branch2.blocks.5.attn.proj.bias", + "backbone.branch2.blocks.5.norm2.weight", + "backbone.branch2.blocks.5.norm2.bias", + "backbone.branch2.blocks.5.mlp.fc1.bias", + "backbone.branch2.blocks.5.mlp.fc2.bias", + "backbone.branch3.blocks.5.gamma_1", + "backbone.branch3.blocks.5.gamma_2", + "backbone.branch3.blocks.5.norm1.weight", + "backbone.branch3.blocks.5.norm1.bias", + "backbone.branch3.blocks.5.attn.qkv.bias", + "backbone.branch3.blocks.5.attn.proj.bias", + "backbone.branch3.blocks.5.norm2.weight", + "backbone.branch3.blocks.5.norm2.bias", + "backbone.branch3.blocks.5.mlp.fc1.bias", + "backbone.branch3.blocks.5.mlp.fc2.bias" + ], + "lr_scale": 0.12090549356574626, + "lr": 1.2090549356574626e-05, + "weight_decay": 0.0 + }, + "layer_12_decay": { + "param_names": [ + "backbone.branch1.blocks.11.attn.relative_position_bias_table", + "backbone.branch1.blocks.11.attn.qkv.weight", + "backbone.branch1.blocks.11.attn.proj.weight", + "backbone.branch1.blocks.11.mlp.fc1.weight", + "backbone.branch1.blocks.11.mlp.fc2.weight", + "backbone.branch2.blocks.5.attn.relative_position_bias_table", + "backbone.branch2.blocks.5.attn.qkv.weight", + "backbone.branch2.blocks.5.attn.proj.weight", + "backbone.branch2.blocks.5.mlp.fc1.weight", + "backbone.branch2.blocks.5.mlp.fc2.weight", + "backbone.branch3.blocks.5.attn.qkv.weight", + "backbone.branch3.blocks.5.attn.proj.weight", + "backbone.branch3.blocks.5.mlp.fc1.weight", + "backbone.branch3.blocks.5.mlp.fc2.weight" + ], + "lr_scale": 0.12090549356574626, + "lr": 1.2090549356574626e-05, + "weight_decay": 0.05 + }, + "layer_13_no_decay": { + "param_names": [ + "backbone.branch1.blocks.12.gamma_1", + "backbone.branch1.blocks.12.gamma_2", + "backbone.branch1.blocks.12.norm1.weight", + "backbone.branch1.blocks.12.norm1.bias", + "backbone.branch1.blocks.12.attn.q_bias", + "backbone.branch1.blocks.12.attn.v_bias", + "backbone.branch1.blocks.12.attn.proj.bias", + "backbone.branch1.blocks.12.norm2.weight", + "backbone.branch1.blocks.12.norm2.bias", + "backbone.branch1.blocks.12.mlp.fc1.bias", + "backbone.branch1.blocks.12.mlp.fc2.bias" + ], + "lr_scale": 0.14224175713617207, + "lr": 1.4224175713617208e-05, + "weight_decay": 0.0 + }, + "layer_13_decay": { + "param_names": [ + "backbone.branch1.blocks.12.attn.relative_position_bias_table", + "backbone.branch1.blocks.12.attn.qkv.weight", + "backbone.branch1.blocks.12.attn.proj.weight", + "backbone.branch1.blocks.12.mlp.fc1.weight", + "backbone.branch1.blocks.12.mlp.fc2.weight" + ], + "lr_scale": 0.14224175713617207, + "lr": 1.4224175713617208e-05, + "weight_decay": 0.05 + }, + "layer_14_no_decay": { + "param_names": [ + "backbone.branch1.blocks.13.gamma_1", + "backbone.branch1.blocks.13.gamma_2", + "backbone.branch1.blocks.13.norm1.weight", + "backbone.branch1.blocks.13.norm1.bias", + "backbone.branch1.blocks.13.attn.q_bias", + "backbone.branch1.blocks.13.attn.v_bias", + "backbone.branch1.blocks.13.attn.proj.bias", + "backbone.branch1.blocks.13.norm2.weight", + "backbone.branch1.blocks.13.norm2.bias", + "backbone.branch1.blocks.13.mlp.fc1.bias", + "backbone.branch1.blocks.13.mlp.fc2.bias", + "backbone.branch2.blocks.6.gamma_1", + "backbone.branch2.blocks.6.gamma_2", + "backbone.branch2.blocks.6.norm1.weight", + "backbone.branch2.blocks.6.norm1.bias", + "backbone.branch2.blocks.6.attn.q_bias", + "backbone.branch2.blocks.6.attn.v_bias", + "backbone.branch2.blocks.6.attn.proj.bias", + "backbone.branch2.blocks.6.norm2.weight", + "backbone.branch2.blocks.6.norm2.bias", + "backbone.branch2.blocks.6.mlp.fc1.bias", + "backbone.branch2.blocks.6.mlp.fc2.bias", + "backbone.branch3.blocks.6.gamma_1", + "backbone.branch3.blocks.6.gamma_2", + "backbone.branch3.blocks.6.norm1.weight", + "backbone.branch3.blocks.6.norm1.bias", + "backbone.branch3.blocks.6.attn.qkv.bias", + "backbone.branch3.blocks.6.attn.proj.bias", + "backbone.branch3.blocks.6.norm2.weight", + "backbone.branch3.blocks.6.norm2.bias", + "backbone.branch3.blocks.6.mlp.fc1.bias", + "backbone.branch3.blocks.6.mlp.fc2.bias" + ], + "lr_scale": 0.1673432436896142, + "lr": 1.673432436896142e-05, + "weight_decay": 0.0 + }, + "layer_14_decay": { + "param_names": [ + "backbone.branch1.blocks.13.attn.relative_position_bias_table", + "backbone.branch1.blocks.13.attn.qkv.weight", + "backbone.branch1.blocks.13.attn.proj.weight", + "backbone.branch1.blocks.13.mlp.fc1.weight", + "backbone.branch1.blocks.13.mlp.fc2.weight", + "backbone.branch2.blocks.6.attn.relative_position_bias_table", + "backbone.branch2.blocks.6.attn.qkv.weight", + "backbone.branch2.blocks.6.attn.proj.weight", + "backbone.branch2.blocks.6.mlp.fc1.weight", + "backbone.branch2.blocks.6.mlp.fc2.weight", + "backbone.branch3.blocks.6.attn.qkv.weight", + "backbone.branch3.blocks.6.attn.proj.weight", + "backbone.branch3.blocks.6.mlp.fc1.weight", + "backbone.branch3.blocks.6.mlp.fc2.weight" + ], + "lr_scale": 0.1673432436896142, + "lr": 1.673432436896142e-05, + "weight_decay": 0.05 + }, + "layer_15_no_decay": { + "param_names": [ + "backbone.branch1.blocks.14.gamma_1", + "backbone.branch1.blocks.14.gamma_2", + "backbone.branch1.blocks.14.norm1.weight", + "backbone.branch1.blocks.14.norm1.bias", + "backbone.branch1.blocks.14.attn.q_bias", + "backbone.branch1.blocks.14.attn.v_bias", + "backbone.branch1.blocks.14.attn.proj.bias", + "backbone.branch1.blocks.14.norm2.weight", + "backbone.branch1.blocks.14.norm2.bias", + "backbone.branch1.blocks.14.mlp.fc1.bias", + "backbone.branch1.blocks.14.mlp.fc2.bias" + ], + "lr_scale": 0.1968744043407226, + "lr": 1.968744043407226e-05, + "weight_decay": 0.0 + }, + "layer_15_decay": { + "param_names": [ + "backbone.branch1.blocks.14.attn.relative_position_bias_table", + "backbone.branch1.blocks.14.attn.qkv.weight", + "backbone.branch1.blocks.14.attn.proj.weight", + "backbone.branch1.blocks.14.mlp.fc1.weight", + "backbone.branch1.blocks.14.mlp.fc2.weight" + ], + "lr_scale": 0.1968744043407226, + "lr": 1.968744043407226e-05, + "weight_decay": 0.05 + }, + "layer_16_no_decay": { + "param_names": [ + "backbone.branch1.blocks.15.gamma_1", + "backbone.branch1.blocks.15.gamma_2", + "backbone.branch1.blocks.15.norm1.weight", + "backbone.branch1.blocks.15.norm1.bias", + "backbone.branch1.blocks.15.attn.q_bias", + "backbone.branch1.blocks.15.attn.v_bias", + "backbone.branch1.blocks.15.attn.proj.bias", + "backbone.branch1.blocks.15.norm2.weight", + "backbone.branch1.blocks.15.norm2.bias", + "backbone.branch1.blocks.15.mlp.fc1.bias", + "backbone.branch1.blocks.15.mlp.fc2.bias", + "backbone.branch2.blocks.7.gamma_1", + "backbone.branch2.blocks.7.gamma_2", + "backbone.branch2.blocks.7.norm1.weight", + "backbone.branch2.blocks.7.norm1.bias", + "backbone.branch2.blocks.7.attn.q_bias", + "backbone.branch2.blocks.7.attn.v_bias", + "backbone.branch2.blocks.7.attn.proj.bias", + "backbone.branch2.blocks.7.norm2.weight", + "backbone.branch2.blocks.7.norm2.bias", + "backbone.branch2.blocks.7.mlp.fc1.bias", + "backbone.branch2.blocks.7.mlp.fc2.bias", + "backbone.branch3.blocks.7.gamma_1", + "backbone.branch3.blocks.7.gamma_2", + "backbone.branch3.blocks.7.norm1.weight", + "backbone.branch3.blocks.7.norm1.bias", + "backbone.branch3.blocks.7.attn.qkv.bias", + "backbone.branch3.blocks.7.attn.proj.bias", + "backbone.branch3.blocks.7.norm2.weight", + "backbone.branch3.blocks.7.norm2.bias", + "backbone.branch3.blocks.7.mlp.fc1.bias", + "backbone.branch3.blocks.7.mlp.fc2.bias" + ], + "lr_scale": 0.23161694628320306, + "lr": 2.3161694628320308e-05, + "weight_decay": 0.0 + }, + "layer_16_decay": { + "param_names": [ + "backbone.branch1.blocks.15.attn.relative_position_bias_table", + "backbone.branch1.blocks.15.attn.qkv.weight", + "backbone.branch1.blocks.15.attn.proj.weight", + "backbone.branch1.blocks.15.mlp.fc1.weight", + "backbone.branch1.blocks.15.mlp.fc2.weight", + "backbone.branch2.blocks.7.attn.relative_position_bias_table", + "backbone.branch2.blocks.7.attn.qkv.weight", + "backbone.branch2.blocks.7.attn.proj.weight", + "backbone.branch2.blocks.7.mlp.fc1.weight", + "backbone.branch2.blocks.7.mlp.fc2.weight", + "backbone.branch3.blocks.7.attn.qkv.weight", + "backbone.branch3.blocks.7.attn.proj.weight", + "backbone.branch3.blocks.7.mlp.fc1.weight", + "backbone.branch3.blocks.7.mlp.fc2.weight" + ], + "lr_scale": 0.23161694628320306, + "lr": 2.3161694628320308e-05, + "weight_decay": 0.05 + }, + "layer_17_no_decay": { + "param_names": [ + "backbone.branch1.blocks.16.gamma_1", + "backbone.branch1.blocks.16.gamma_2", + "backbone.branch1.blocks.16.norm1.weight", + "backbone.branch1.blocks.16.norm1.bias", + "backbone.branch1.blocks.16.attn.q_bias", + "backbone.branch1.blocks.16.attn.v_bias", + "backbone.branch1.blocks.16.attn.proj.bias", + "backbone.branch1.blocks.16.norm2.weight", + "backbone.branch1.blocks.16.norm2.bias", + "backbone.branch1.blocks.16.mlp.fc1.bias", + "backbone.branch1.blocks.16.mlp.fc2.bias" + ], + "lr_scale": 0.27249052503906246, + "lr": 2.7249052503906248e-05, + "weight_decay": 0.0 + }, + "layer_17_decay": { + "param_names": [ + "backbone.branch1.blocks.16.attn.relative_position_bias_table", + "backbone.branch1.blocks.16.attn.qkv.weight", + "backbone.branch1.blocks.16.attn.proj.weight", + "backbone.branch1.blocks.16.mlp.fc1.weight", + "backbone.branch1.blocks.16.mlp.fc2.weight" + ], + "lr_scale": 0.27249052503906246, + "lr": 2.7249052503906248e-05, + "weight_decay": 0.05 + }, + "layer_18_no_decay": { + "param_names": [ + "backbone.branch1.blocks.17.gamma_1", + "backbone.branch1.blocks.17.gamma_2", + "backbone.branch1.blocks.17.norm1.weight", + "backbone.branch1.blocks.17.norm1.bias", + "backbone.branch1.blocks.17.attn.q_bias", + "backbone.branch1.blocks.17.attn.v_bias", + "backbone.branch1.blocks.17.attn.proj.bias", + "backbone.branch1.blocks.17.norm2.weight", + "backbone.branch1.blocks.17.norm2.bias", + "backbone.branch1.blocks.17.mlp.fc1.bias", + "backbone.branch1.blocks.17.mlp.fc2.bias", + "backbone.branch2.blocks.8.gamma_1", + "backbone.branch2.blocks.8.gamma_2", + "backbone.branch2.blocks.8.norm1.weight", + "backbone.branch2.blocks.8.norm1.bias", + "backbone.branch2.blocks.8.attn.q_bias", + "backbone.branch2.blocks.8.attn.v_bias", + "backbone.branch2.blocks.8.attn.proj.bias", + "backbone.branch2.blocks.8.norm2.weight", + "backbone.branch2.blocks.8.norm2.bias", + "backbone.branch2.blocks.8.mlp.fc1.bias", + "backbone.branch2.blocks.8.mlp.fc2.bias", + "backbone.branch3.blocks.8.gamma_1", + "backbone.branch3.blocks.8.gamma_2", + "backbone.branch3.blocks.8.norm1.weight", + "backbone.branch3.blocks.8.norm1.bias", + "backbone.branch3.blocks.8.attn.qkv.bias", + "backbone.branch3.blocks.8.attn.proj.bias", + "backbone.branch3.blocks.8.norm2.weight", + "backbone.branch3.blocks.8.norm2.bias", + "backbone.branch3.blocks.8.mlp.fc1.bias", + "backbone.branch3.blocks.8.mlp.fc2.bias" + ], + "lr_scale": 0.3205770882812499, + "lr": 3.2057708828124995e-05, + "weight_decay": 0.0 + }, + "layer_18_decay": { + "param_names": [ + "backbone.branch1.blocks.17.attn.relative_position_bias_table", + "backbone.branch1.blocks.17.attn.qkv.weight", + "backbone.branch1.blocks.17.attn.proj.weight", + "backbone.branch1.blocks.17.mlp.fc1.weight", + "backbone.branch1.blocks.17.mlp.fc2.weight", + "backbone.branch2.blocks.8.attn.relative_position_bias_table", + "backbone.branch2.blocks.8.attn.qkv.weight", + "backbone.branch2.blocks.8.attn.proj.weight", + "backbone.branch2.blocks.8.mlp.fc1.weight", + "backbone.branch2.blocks.8.mlp.fc2.weight", + "backbone.branch3.blocks.8.attn.qkv.weight", + "backbone.branch3.blocks.8.attn.proj.weight", + "backbone.branch3.blocks.8.mlp.fc1.weight", + "backbone.branch3.blocks.8.mlp.fc2.weight" + ], + "lr_scale": 0.3205770882812499, + "lr": 3.2057708828124995e-05, + "weight_decay": 0.05 + }, + "layer_19_no_decay": { + "param_names": [ + "backbone.branch1.blocks.18.gamma_1", + "backbone.branch1.blocks.18.gamma_2", + "backbone.branch1.blocks.18.norm1.weight", + "backbone.branch1.blocks.18.norm1.bias", + "backbone.branch1.blocks.18.attn.q_bias", + "backbone.branch1.blocks.18.attn.v_bias", + "backbone.branch1.blocks.18.attn.proj.bias", + "backbone.branch1.blocks.18.norm2.weight", + "backbone.branch1.blocks.18.norm2.bias", + "backbone.branch1.blocks.18.mlp.fc1.bias", + "backbone.branch1.blocks.18.mlp.fc2.bias" + ], + "lr_scale": 0.37714951562499993, + "lr": 3.77149515625e-05, + "weight_decay": 0.0 + }, + "layer_19_decay": { + "param_names": [ + "backbone.branch1.blocks.18.attn.relative_position_bias_table", + "backbone.branch1.blocks.18.attn.qkv.weight", + "backbone.branch1.blocks.18.attn.proj.weight", + "backbone.branch1.blocks.18.mlp.fc1.weight", + "backbone.branch1.blocks.18.mlp.fc2.weight" + ], + "lr_scale": 0.37714951562499993, + "lr": 3.77149515625e-05, + "weight_decay": 0.05 + }, + "layer_20_no_decay": { + "param_names": [ + "backbone.branch1.blocks.19.gamma_1", + "backbone.branch1.blocks.19.gamma_2", + "backbone.branch1.blocks.19.norm1.weight", + "backbone.branch1.blocks.19.norm1.bias", + "backbone.branch1.blocks.19.attn.q_bias", + "backbone.branch1.blocks.19.attn.v_bias", + "backbone.branch1.blocks.19.attn.proj.bias", + "backbone.branch1.blocks.19.norm2.weight", + "backbone.branch1.blocks.19.norm2.bias", + "backbone.branch1.blocks.19.mlp.fc1.bias", + "backbone.branch1.blocks.19.mlp.fc2.bias", + "backbone.branch2.blocks.9.gamma_1", + "backbone.branch2.blocks.9.gamma_2", + "backbone.branch2.blocks.9.norm1.weight", + "backbone.branch2.blocks.9.norm1.bias", + "backbone.branch2.blocks.9.attn.q_bias", + "backbone.branch2.blocks.9.attn.v_bias", + "backbone.branch2.blocks.9.attn.proj.bias", + "backbone.branch2.blocks.9.norm2.weight", + "backbone.branch2.blocks.9.norm2.bias", + "backbone.branch2.blocks.9.mlp.fc1.bias", + "backbone.branch2.blocks.9.mlp.fc2.bias", + "backbone.branch3.blocks.9.gamma_1", + "backbone.branch3.blocks.9.gamma_2", + "backbone.branch3.blocks.9.norm1.weight", + "backbone.branch3.blocks.9.norm1.bias", + "backbone.branch3.blocks.9.attn.qkv.bias", + "backbone.branch3.blocks.9.attn.proj.bias", + "backbone.branch3.blocks.9.norm2.weight", + "backbone.branch3.blocks.9.norm2.bias", + "backbone.branch3.blocks.9.mlp.fc1.bias", + "backbone.branch3.blocks.9.mlp.fc2.bias" + ], + "lr_scale": 0.44370531249999995, + "lr": 4.4370531249999995e-05, + "weight_decay": 0.0 + }, + "layer_20_decay": { + "param_names": [ + "backbone.branch1.blocks.19.attn.relative_position_bias_table", + "backbone.branch1.blocks.19.attn.qkv.weight", + "backbone.branch1.blocks.19.attn.proj.weight", + "backbone.branch1.blocks.19.mlp.fc1.weight", + "backbone.branch1.blocks.19.mlp.fc2.weight", + "backbone.branch2.blocks.9.attn.relative_position_bias_table", + "backbone.branch2.blocks.9.attn.qkv.weight", + "backbone.branch2.blocks.9.attn.proj.weight", + "backbone.branch2.blocks.9.mlp.fc1.weight", + "backbone.branch2.blocks.9.mlp.fc2.weight", + "backbone.branch3.blocks.9.attn.qkv.weight", + "backbone.branch3.blocks.9.attn.proj.weight", + "backbone.branch3.blocks.9.mlp.fc1.weight", + "backbone.branch3.blocks.9.mlp.fc2.weight" + ], + "lr_scale": 0.44370531249999995, + "lr": 4.4370531249999995e-05, + "weight_decay": 0.05 + }, + "layer_21_no_decay": { + "param_names": [ + "backbone.branch1.blocks.20.gamma_1", + "backbone.branch1.blocks.20.gamma_2", + "backbone.branch1.blocks.20.norm1.weight", + "backbone.branch1.blocks.20.norm1.bias", + "backbone.branch1.blocks.20.attn.q_bias", + "backbone.branch1.blocks.20.attn.v_bias", + "backbone.branch1.blocks.20.attn.proj.bias", + "backbone.branch1.blocks.20.norm2.weight", + "backbone.branch1.blocks.20.norm2.bias", + "backbone.branch1.blocks.20.mlp.fc1.bias", + "backbone.branch1.blocks.20.mlp.fc2.bias" + ], + "lr_scale": 0.5220062499999999, + "lr": 5.220062499999999e-05, + "weight_decay": 0.0 + }, + "layer_21_decay": { + "param_names": [ + "backbone.branch1.blocks.20.attn.relative_position_bias_table", + "backbone.branch1.blocks.20.attn.qkv.weight", + "backbone.branch1.blocks.20.attn.proj.weight", + "backbone.branch1.blocks.20.mlp.fc1.weight", + "backbone.branch1.blocks.20.mlp.fc2.weight" + ], + "lr_scale": 0.5220062499999999, + "lr": 5.220062499999999e-05, + "weight_decay": 0.05 + }, + "layer_22_no_decay": { + "param_names": [ + "backbone.branch1.blocks.21.gamma_1", + "backbone.branch1.blocks.21.gamma_2", + "backbone.branch1.blocks.21.norm1.weight", + "backbone.branch1.blocks.21.norm1.bias", + "backbone.branch1.blocks.21.attn.q_bias", + "backbone.branch1.blocks.21.attn.v_bias", + "backbone.branch1.blocks.21.attn.proj.bias", + "backbone.branch1.blocks.21.norm2.weight", + "backbone.branch1.blocks.21.norm2.bias", + "backbone.branch1.blocks.21.mlp.fc1.bias", + "backbone.branch1.blocks.21.mlp.fc2.bias", + "backbone.branch2.blocks.10.gamma_1", + "backbone.branch2.blocks.10.gamma_2", + "backbone.branch2.blocks.10.norm1.weight", + "backbone.branch2.blocks.10.norm1.bias", + "backbone.branch2.blocks.10.attn.q_bias", + "backbone.branch2.blocks.10.attn.v_bias", + "backbone.branch2.blocks.10.attn.proj.bias", + "backbone.branch2.blocks.10.norm2.weight", + "backbone.branch2.blocks.10.norm2.bias", + "backbone.branch2.blocks.10.mlp.fc1.bias", + "backbone.branch2.blocks.10.mlp.fc2.bias", + "backbone.branch3.blocks.10.gamma_1", + "backbone.branch3.blocks.10.gamma_2", + "backbone.branch3.blocks.10.norm1.weight", + "backbone.branch3.blocks.10.norm1.bias", + "backbone.branch3.blocks.10.attn.qkv.bias", + "backbone.branch3.blocks.10.attn.proj.bias", + "backbone.branch3.blocks.10.norm2.weight", + "backbone.branch3.blocks.10.norm2.bias", + "backbone.branch3.blocks.10.mlp.fc1.bias", + "backbone.branch3.blocks.10.mlp.fc2.bias" + ], + "lr_scale": 0.6141249999999999, + "lr": 6.14125e-05, + "weight_decay": 0.0 + }, + "layer_22_decay": { + "param_names": [ + "backbone.branch1.blocks.21.attn.relative_position_bias_table", + "backbone.branch1.blocks.21.attn.qkv.weight", + "backbone.branch1.blocks.21.attn.proj.weight", + "backbone.branch1.blocks.21.mlp.fc1.weight", + "backbone.branch1.blocks.21.mlp.fc2.weight", + "backbone.branch2.blocks.10.attn.relative_position_bias_table", + "backbone.branch2.blocks.10.attn.qkv.weight", + "backbone.branch2.blocks.10.attn.proj.weight", + "backbone.branch2.blocks.10.mlp.fc1.weight", + "backbone.branch2.blocks.10.mlp.fc2.weight", + "backbone.branch3.blocks.10.attn.qkv.weight", + "backbone.branch3.blocks.10.attn.proj.weight", + "backbone.branch3.blocks.10.mlp.fc1.weight", + "backbone.branch3.blocks.10.mlp.fc2.weight" + ], + "lr_scale": 0.6141249999999999, + "lr": 6.14125e-05, + "weight_decay": 0.05 + }, + "layer_23_no_decay": { + "param_names": [ + "backbone.branch1.blocks.22.gamma_1", + "backbone.branch1.blocks.22.gamma_2", + "backbone.branch1.blocks.22.norm1.weight", + "backbone.branch1.blocks.22.norm1.bias", + "backbone.branch1.blocks.22.attn.q_bias", + "backbone.branch1.blocks.22.attn.v_bias", + "backbone.branch1.blocks.22.attn.proj.bias", + "backbone.branch1.blocks.22.norm2.weight", + "backbone.branch1.blocks.22.norm2.bias", + "backbone.branch1.blocks.22.mlp.fc1.bias", + "backbone.branch1.blocks.22.mlp.fc2.bias" + ], + "lr_scale": 0.7224999999999999, + "lr": 7.225e-05, + "weight_decay": 0.0 + }, + "layer_23_decay": { + "param_names": [ + "backbone.branch1.blocks.22.attn.relative_position_bias_table", + "backbone.branch1.blocks.22.attn.qkv.weight", + "backbone.branch1.blocks.22.attn.proj.weight", + "backbone.branch1.blocks.22.mlp.fc1.weight", + "backbone.branch1.blocks.22.mlp.fc2.weight" + ], + "lr_scale": 0.7224999999999999, + "lr": 7.225e-05, + "weight_decay": 0.05 + }, + "layer_24_no_decay": { + "param_names": [ + "backbone.branch1.blocks.23.gamma_1", + "backbone.branch1.blocks.23.gamma_2", + "backbone.branch1.blocks.23.norm1.weight", + "backbone.branch1.blocks.23.norm1.bias", + "backbone.branch1.blocks.23.attn.q_bias", + "backbone.branch1.blocks.23.attn.v_bias", + "backbone.branch1.blocks.23.attn.proj.bias", + "backbone.branch1.blocks.23.norm2.weight", + "backbone.branch1.blocks.23.norm2.bias", + "backbone.branch1.blocks.23.mlp.fc1.bias", + "backbone.branch1.blocks.23.mlp.fc2.bias", + "backbone.branch2.blocks.11.gamma_1", + "backbone.branch2.blocks.11.gamma_2", + "backbone.branch2.blocks.11.norm1.weight", + "backbone.branch2.blocks.11.norm1.bias", + "backbone.branch2.blocks.11.attn.q_bias", + "backbone.branch2.blocks.11.attn.v_bias", + "backbone.branch2.blocks.11.attn.proj.bias", + "backbone.branch2.blocks.11.norm2.weight", + "backbone.branch2.blocks.11.norm2.bias", + "backbone.branch2.blocks.11.mlp.fc1.bias", + "backbone.branch2.blocks.11.mlp.fc2.bias", + "backbone.branch3.blocks.11.gamma_1", + "backbone.branch3.blocks.11.gamma_2", + "backbone.branch3.blocks.11.norm1.weight", + "backbone.branch3.blocks.11.norm1.bias", + "backbone.branch3.blocks.11.attn.qkv.bias", + "backbone.branch3.blocks.11.attn.proj.bias", + "backbone.branch3.blocks.11.norm2.weight", + "backbone.branch3.blocks.11.norm2.bias", + "backbone.branch3.blocks.11.mlp.fc1.bias", + "backbone.branch3.blocks.11.mlp.fc2.bias" + ], + "lr_scale": 0.85, + "lr": 8.5e-05, + "weight_decay": 0.0 + }, + "layer_24_decay": { + "param_names": [ + "backbone.branch1.blocks.23.attn.relative_position_bias_table", + "backbone.branch1.blocks.23.attn.qkv.weight", + "backbone.branch1.blocks.23.attn.proj.weight", + "backbone.branch1.blocks.23.mlp.fc1.weight", + "backbone.branch1.blocks.23.mlp.fc2.weight", + "backbone.branch2.blocks.11.attn.relative_position_bias_table", + "backbone.branch2.blocks.11.attn.qkv.weight", + "backbone.branch2.blocks.11.attn.proj.weight", + "backbone.branch2.blocks.11.mlp.fc1.weight", + "backbone.branch2.blocks.11.mlp.fc2.weight", + "backbone.branch3.blocks.11.attn.qkv.weight", + "backbone.branch3.blocks.11.attn.proj.weight", + "backbone.branch3.blocks.11.mlp.fc1.weight", + "backbone.branch3.blocks.11.mlp.fc2.weight" + ], + "lr_scale": 0.85, + "lr": 8.5e-05, + "weight_decay": 0.05 + }, + "layer_25_no_decay": { + "param_names": [ + "backbone.interactions.0.interaction_units_12.branch2to1_proj.bias", + "backbone.interactions.0.interaction_units_12.branch2to1_injector.ca_gamma", + "backbone.interactions.0.interaction_units_12.branch2to1_injector.cffn_gamma", + "backbone.interactions.0.interaction_units_12.branch2to1_injector.query_norm.weight", + "backbone.interactions.0.interaction_units_12.branch2to1_injector.query_norm.bias", + "backbone.interactions.0.interaction_units_12.branch2to1_injector.feat_norm.weight", + "backbone.interactions.0.interaction_units_12.branch2to1_injector.feat_norm.bias", + "backbone.interactions.0.interaction_units_12.branch2to1_injector.attn.sampling_offsets.bias", + "backbone.interactions.0.interaction_units_12.branch2to1_injector.attn.attention_weights.bias", + "backbone.interactions.0.interaction_units_12.branch2to1_injector.attn.value_proj.bias", + "backbone.interactions.0.interaction_units_12.branch2to1_injector.attn.output_proj.bias", + "backbone.interactions.0.interaction_units_12.branch2to1_injector.ffn.fc1.bias", + "backbone.interactions.0.interaction_units_12.branch2to1_injector.ffn.dwconv.dwconv.bias", + "backbone.interactions.0.interaction_units_12.branch2to1_injector.ffn.fc2.bias", + "backbone.interactions.0.interaction_units_12.branch2to1_injector.ffn_norm.weight", + "backbone.interactions.0.interaction_units_12.branch2to1_injector.ffn_norm.bias", + "backbone.interactions.0.interaction_units_12.branch1to2_proj.bias", + "backbone.interactions.0.interaction_units_12.branch1to2_injector.ca_gamma", + "backbone.interactions.0.interaction_units_12.branch1to2_injector.cffn_gamma", + "backbone.interactions.0.interaction_units_12.branch1to2_injector.query_norm.weight", + "backbone.interactions.0.interaction_units_12.branch1to2_injector.query_norm.bias", + "backbone.interactions.0.interaction_units_12.branch1to2_injector.feat_norm.weight", + "backbone.interactions.0.interaction_units_12.branch1to2_injector.feat_norm.bias", + "backbone.interactions.0.interaction_units_12.branch1to2_injector.attn.sampling_offsets.bias", + "backbone.interactions.0.interaction_units_12.branch1to2_injector.attn.attention_weights.bias", + "backbone.interactions.0.interaction_units_12.branch1to2_injector.attn.value_proj.bias", + "backbone.interactions.0.interaction_units_12.branch1to2_injector.attn.output_proj.bias", + "backbone.interactions.0.interaction_units_12.branch1to2_injector.ffn.fc1.bias", + "backbone.interactions.0.interaction_units_12.branch1to2_injector.ffn.dwconv.dwconv.bias", + "backbone.interactions.0.interaction_units_12.branch1to2_injector.ffn.fc2.bias", + "backbone.interactions.0.interaction_units_12.branch1to2_injector.ffn_norm.weight", + "backbone.interactions.0.interaction_units_12.branch1to2_injector.ffn_norm.bias", + "backbone.interactions.0.interaction_units_23.branch2to1_proj.bias", + "backbone.interactions.0.interaction_units_23.branch2to1_injector.ca_gamma", + "backbone.interactions.0.interaction_units_23.branch2to1_injector.cffn_gamma", + "backbone.interactions.0.interaction_units_23.branch2to1_injector.query_norm.weight", + "backbone.interactions.0.interaction_units_23.branch2to1_injector.query_norm.bias", + "backbone.interactions.0.interaction_units_23.branch2to1_injector.feat_norm.weight", + "backbone.interactions.0.interaction_units_23.branch2to1_injector.feat_norm.bias", + "backbone.interactions.0.interaction_units_23.branch2to1_injector.attn.sampling_offsets.bias", + "backbone.interactions.0.interaction_units_23.branch2to1_injector.attn.attention_weights.bias", + "backbone.interactions.0.interaction_units_23.branch2to1_injector.attn.value_proj.bias", + "backbone.interactions.0.interaction_units_23.branch2to1_injector.attn.output_proj.bias", + "backbone.interactions.0.interaction_units_23.branch2to1_injector.ffn.fc1.bias", + "backbone.interactions.0.interaction_units_23.branch2to1_injector.ffn.dwconv.dwconv.bias", + "backbone.interactions.0.interaction_units_23.branch2to1_injector.ffn.fc2.bias", + "backbone.interactions.0.interaction_units_23.branch2to1_injector.ffn_norm.weight", + "backbone.interactions.0.interaction_units_23.branch2to1_injector.ffn_norm.bias", + "backbone.interactions.0.interaction_units_23.branch1to2_proj.bias", + "backbone.interactions.0.interaction_units_23.branch1to2_injector.ca_gamma", + "backbone.interactions.0.interaction_units_23.branch1to2_injector.cffn_gamma", + "backbone.interactions.0.interaction_units_23.branch1to2_injector.query_norm.weight", + "backbone.interactions.0.interaction_units_23.branch1to2_injector.query_norm.bias", + "backbone.interactions.0.interaction_units_23.branch1to2_injector.feat_norm.weight", + "backbone.interactions.0.interaction_units_23.branch1to2_injector.feat_norm.bias", + "backbone.interactions.0.interaction_units_23.branch1to2_injector.attn.sampling_offsets.bias", + "backbone.interactions.0.interaction_units_23.branch1to2_injector.attn.attention_weights.bias", + "backbone.interactions.0.interaction_units_23.branch1to2_injector.attn.value_proj.bias", + "backbone.interactions.0.interaction_units_23.branch1to2_injector.attn.output_proj.bias", + "backbone.interactions.0.interaction_units_23.branch1to2_injector.ffn.fc1.bias", + "backbone.interactions.0.interaction_units_23.branch1to2_injector.ffn.dwconv.dwconv.bias", + "backbone.interactions.0.interaction_units_23.branch1to2_injector.ffn.fc2.bias", + "backbone.interactions.0.interaction_units_23.branch1to2_injector.ffn_norm.weight", + "backbone.interactions.0.interaction_units_23.branch1to2_injector.ffn_norm.bias", + "backbone.interactions.1.interaction_units_12.branch2to1_proj.bias", + "backbone.interactions.1.interaction_units_12.branch2to1_injector.ca_gamma", + "backbone.interactions.1.interaction_units_12.branch2to1_injector.cffn_gamma", + "backbone.interactions.1.interaction_units_12.branch2to1_injector.query_norm.weight", + "backbone.interactions.1.interaction_units_12.branch2to1_injector.query_norm.bias", + "backbone.interactions.1.interaction_units_12.branch2to1_injector.feat_norm.weight", + "backbone.interactions.1.interaction_units_12.branch2to1_injector.feat_norm.bias", + "backbone.interactions.1.interaction_units_12.branch2to1_injector.attn.sampling_offsets.bias", + "backbone.interactions.1.interaction_units_12.branch2to1_injector.attn.attention_weights.bias", + "backbone.interactions.1.interaction_units_12.branch2to1_injector.attn.value_proj.bias", + "backbone.interactions.1.interaction_units_12.branch2to1_injector.attn.output_proj.bias", + "backbone.interactions.1.interaction_units_12.branch2to1_injector.ffn.fc1.bias", + "backbone.interactions.1.interaction_units_12.branch2to1_injector.ffn.dwconv.dwconv.bias", + "backbone.interactions.1.interaction_units_12.branch2to1_injector.ffn.fc2.bias", + "backbone.interactions.1.interaction_units_12.branch2to1_injector.ffn_norm.weight", + "backbone.interactions.1.interaction_units_12.branch2to1_injector.ffn_norm.bias", + "backbone.interactions.1.interaction_units_12.branch1to2_proj.bias", + "backbone.interactions.1.interaction_units_12.branch1to2_injector.ca_gamma", + "backbone.interactions.1.interaction_units_12.branch1to2_injector.cffn_gamma", + "backbone.interactions.1.interaction_units_12.branch1to2_injector.query_norm.weight", + "backbone.interactions.1.interaction_units_12.branch1to2_injector.query_norm.bias", + "backbone.interactions.1.interaction_units_12.branch1to2_injector.feat_norm.weight", + "backbone.interactions.1.interaction_units_12.branch1to2_injector.feat_norm.bias", + "backbone.interactions.1.interaction_units_12.branch1to2_injector.attn.sampling_offsets.bias", + "backbone.interactions.1.interaction_units_12.branch1to2_injector.attn.attention_weights.bias", + "backbone.interactions.1.interaction_units_12.branch1to2_injector.attn.value_proj.bias", + "backbone.interactions.1.interaction_units_12.branch1to2_injector.attn.output_proj.bias", + "backbone.interactions.1.interaction_units_12.branch1to2_injector.ffn.fc1.bias", + "backbone.interactions.1.interaction_units_12.branch1to2_injector.ffn.dwconv.dwconv.bias", + "backbone.interactions.1.interaction_units_12.branch1to2_injector.ffn.fc2.bias", + "backbone.interactions.1.interaction_units_12.branch1to2_injector.ffn_norm.weight", + "backbone.interactions.1.interaction_units_12.branch1to2_injector.ffn_norm.bias", + "backbone.interactions.1.interaction_units_23.branch2to1_proj.bias", + "backbone.interactions.1.interaction_units_23.branch2to1_injector.ca_gamma", + "backbone.interactions.1.interaction_units_23.branch2to1_injector.cffn_gamma", + "backbone.interactions.1.interaction_units_23.branch2to1_injector.query_norm.weight", + "backbone.interactions.1.interaction_units_23.branch2to1_injector.query_norm.bias", + "backbone.interactions.1.interaction_units_23.branch2to1_injector.feat_norm.weight", + "backbone.interactions.1.interaction_units_23.branch2to1_injector.feat_norm.bias", + "backbone.interactions.1.interaction_units_23.branch2to1_injector.attn.sampling_offsets.bias", + "backbone.interactions.1.interaction_units_23.branch2to1_injector.attn.attention_weights.bias", + "backbone.interactions.1.interaction_units_23.branch2to1_injector.attn.value_proj.bias", + "backbone.interactions.1.interaction_units_23.branch2to1_injector.attn.output_proj.bias", + "backbone.interactions.1.interaction_units_23.branch2to1_injector.ffn.fc1.bias", + "backbone.interactions.1.interaction_units_23.branch2to1_injector.ffn.dwconv.dwconv.bias", + "backbone.interactions.1.interaction_units_23.branch2to1_injector.ffn.fc2.bias", + "backbone.interactions.1.interaction_units_23.branch2to1_injector.ffn_norm.weight", + "backbone.interactions.1.interaction_units_23.branch2to1_injector.ffn_norm.bias", + "backbone.interactions.1.interaction_units_23.branch1to2_proj.bias", + "backbone.interactions.1.interaction_units_23.branch1to2_injector.ca_gamma", + "backbone.interactions.1.interaction_units_23.branch1to2_injector.cffn_gamma", + "backbone.interactions.1.interaction_units_23.branch1to2_injector.query_norm.weight", + "backbone.interactions.1.interaction_units_23.branch1to2_injector.query_norm.bias", + "backbone.interactions.1.interaction_units_23.branch1to2_injector.feat_norm.weight", + "backbone.interactions.1.interaction_units_23.branch1to2_injector.feat_norm.bias", + "backbone.interactions.1.interaction_units_23.branch1to2_injector.attn.sampling_offsets.bias", + "backbone.interactions.1.interaction_units_23.branch1to2_injector.attn.attention_weights.bias", + "backbone.interactions.1.interaction_units_23.branch1to2_injector.attn.value_proj.bias", + "backbone.interactions.1.interaction_units_23.branch1to2_injector.attn.output_proj.bias", + "backbone.interactions.1.interaction_units_23.branch1to2_injector.ffn.fc1.bias", + "backbone.interactions.1.interaction_units_23.branch1to2_injector.ffn.dwconv.dwconv.bias", + "backbone.interactions.1.interaction_units_23.branch1to2_injector.ffn.fc2.bias", + "backbone.interactions.1.interaction_units_23.branch1to2_injector.ffn_norm.weight", + "backbone.interactions.1.interaction_units_23.branch1to2_injector.ffn_norm.bias", + "backbone.interactions.2.interaction_units_12.branch2to1_proj.bias", + "backbone.interactions.2.interaction_units_12.branch2to1_injector.ca_gamma", + "backbone.interactions.2.interaction_units_12.branch2to1_injector.cffn_gamma", + "backbone.interactions.2.interaction_units_12.branch2to1_injector.query_norm.weight", + "backbone.interactions.2.interaction_units_12.branch2to1_injector.query_norm.bias", + "backbone.interactions.2.interaction_units_12.branch2to1_injector.feat_norm.weight", + "backbone.interactions.2.interaction_units_12.branch2to1_injector.feat_norm.bias", + "backbone.interactions.2.interaction_units_12.branch2to1_injector.attn.sampling_offsets.bias", + "backbone.interactions.2.interaction_units_12.branch2to1_injector.attn.attention_weights.bias", + "backbone.interactions.2.interaction_units_12.branch2to1_injector.attn.value_proj.bias", + "backbone.interactions.2.interaction_units_12.branch2to1_injector.attn.output_proj.bias", + "backbone.interactions.2.interaction_units_12.branch2to1_injector.ffn.fc1.bias", + "backbone.interactions.2.interaction_units_12.branch2to1_injector.ffn.dwconv.dwconv.bias", + "backbone.interactions.2.interaction_units_12.branch2to1_injector.ffn.fc2.bias", + "backbone.interactions.2.interaction_units_12.branch2to1_injector.ffn_norm.weight", + "backbone.interactions.2.interaction_units_12.branch2to1_injector.ffn_norm.bias", + "backbone.interactions.2.interaction_units_12.branch1to2_proj.bias", + "backbone.interactions.2.interaction_units_12.branch1to2_injector.ca_gamma", + "backbone.interactions.2.interaction_units_12.branch1to2_injector.cffn_gamma", + "backbone.interactions.2.interaction_units_12.branch1to2_injector.query_norm.weight", + "backbone.interactions.2.interaction_units_12.branch1to2_injector.query_norm.bias", + "backbone.interactions.2.interaction_units_12.branch1to2_injector.feat_norm.weight", + "backbone.interactions.2.interaction_units_12.branch1to2_injector.feat_norm.bias", + "backbone.interactions.2.interaction_units_12.branch1to2_injector.attn.sampling_offsets.bias", + "backbone.interactions.2.interaction_units_12.branch1to2_injector.attn.attention_weights.bias", + "backbone.interactions.2.interaction_units_12.branch1to2_injector.attn.value_proj.bias", + "backbone.interactions.2.interaction_units_12.branch1to2_injector.attn.output_proj.bias", + "backbone.interactions.2.interaction_units_12.branch1to2_injector.ffn.fc1.bias", + "backbone.interactions.2.interaction_units_12.branch1to2_injector.ffn.dwconv.dwconv.bias", + "backbone.interactions.2.interaction_units_12.branch1to2_injector.ffn.fc2.bias", + "backbone.interactions.2.interaction_units_12.branch1to2_injector.ffn_norm.weight", + "backbone.interactions.2.interaction_units_12.branch1to2_injector.ffn_norm.bias", + "backbone.interactions.2.interaction_units_23.branch2to1_proj.bias", + "backbone.interactions.2.interaction_units_23.branch2to1_injector.ca_gamma", + "backbone.interactions.2.interaction_units_23.branch2to1_injector.cffn_gamma", + "backbone.interactions.2.interaction_units_23.branch2to1_injector.query_norm.weight", + "backbone.interactions.2.interaction_units_23.branch2to1_injector.query_norm.bias", + "backbone.interactions.2.interaction_units_23.branch2to1_injector.feat_norm.weight", + "backbone.interactions.2.interaction_units_23.branch2to1_injector.feat_norm.bias", + "backbone.interactions.2.interaction_units_23.branch2to1_injector.attn.sampling_offsets.bias", + "backbone.interactions.2.interaction_units_23.branch2to1_injector.attn.attention_weights.bias", + "backbone.interactions.2.interaction_units_23.branch2to1_injector.attn.value_proj.bias", + "backbone.interactions.2.interaction_units_23.branch2to1_injector.attn.output_proj.bias", + "backbone.interactions.2.interaction_units_23.branch2to1_injector.ffn.fc1.bias", + "backbone.interactions.2.interaction_units_23.branch2to1_injector.ffn.dwconv.dwconv.bias", + "backbone.interactions.2.interaction_units_23.branch2to1_injector.ffn.fc2.bias", + "backbone.interactions.2.interaction_units_23.branch2to1_injector.ffn_norm.weight", + "backbone.interactions.2.interaction_units_23.branch2to1_injector.ffn_norm.bias", + "backbone.interactions.2.interaction_units_23.branch1to2_proj.bias", + "backbone.interactions.2.interaction_units_23.branch1to2_injector.ca_gamma", + "backbone.interactions.2.interaction_units_23.branch1to2_injector.cffn_gamma", + "backbone.interactions.2.interaction_units_23.branch1to2_injector.query_norm.weight", + "backbone.interactions.2.interaction_units_23.branch1to2_injector.query_norm.bias", + "backbone.interactions.2.interaction_units_23.branch1to2_injector.feat_norm.weight", + "backbone.interactions.2.interaction_units_23.branch1to2_injector.feat_norm.bias", + "backbone.interactions.2.interaction_units_23.branch1to2_injector.attn.sampling_offsets.bias", + "backbone.interactions.2.interaction_units_23.branch1to2_injector.attn.attention_weights.bias", + "backbone.interactions.2.interaction_units_23.branch1to2_injector.attn.value_proj.bias", + "backbone.interactions.2.interaction_units_23.branch1to2_injector.attn.output_proj.bias", + "backbone.interactions.2.interaction_units_23.branch1to2_injector.ffn.fc1.bias", + "backbone.interactions.2.interaction_units_23.branch1to2_injector.ffn.dwconv.dwconv.bias", + "backbone.interactions.2.interaction_units_23.branch1to2_injector.ffn.fc2.bias", + "backbone.interactions.2.interaction_units_23.branch1to2_injector.ffn_norm.weight", + "backbone.interactions.2.interaction_units_23.branch1to2_injector.ffn_norm.bias", + "backbone.interactions.3.interaction_units_12.branch2to1_proj.bias", + "backbone.interactions.3.interaction_units_12.branch2to1_injector.ca_gamma", + "backbone.interactions.3.interaction_units_12.branch2to1_injector.cffn_gamma", + "backbone.interactions.3.interaction_units_12.branch2to1_injector.query_norm.weight", + "backbone.interactions.3.interaction_units_12.branch2to1_injector.query_norm.bias", + "backbone.interactions.3.interaction_units_12.branch2to1_injector.feat_norm.weight", + "backbone.interactions.3.interaction_units_12.branch2to1_injector.feat_norm.bias", + "backbone.interactions.3.interaction_units_12.branch2to1_injector.attn.sampling_offsets.bias", + "backbone.interactions.3.interaction_units_12.branch2to1_injector.attn.attention_weights.bias", + "backbone.interactions.3.interaction_units_12.branch2to1_injector.attn.value_proj.bias", + "backbone.interactions.3.interaction_units_12.branch2to1_injector.attn.output_proj.bias", + "backbone.interactions.3.interaction_units_12.branch2to1_injector.ffn.fc1.bias", + "backbone.interactions.3.interaction_units_12.branch2to1_injector.ffn.dwconv.dwconv.bias", + "backbone.interactions.3.interaction_units_12.branch2to1_injector.ffn.fc2.bias", + "backbone.interactions.3.interaction_units_12.branch2to1_injector.ffn_norm.weight", + "backbone.interactions.3.interaction_units_12.branch2to1_injector.ffn_norm.bias", + "backbone.interactions.3.interaction_units_12.branch1to2_proj.bias", + "backbone.interactions.3.interaction_units_12.branch1to2_injector.ca_gamma", + "backbone.interactions.3.interaction_units_12.branch1to2_injector.cffn_gamma", + "backbone.interactions.3.interaction_units_12.branch1to2_injector.query_norm.weight", + "backbone.interactions.3.interaction_units_12.branch1to2_injector.query_norm.bias", + "backbone.interactions.3.interaction_units_12.branch1to2_injector.feat_norm.weight", + "backbone.interactions.3.interaction_units_12.branch1to2_injector.feat_norm.bias", + "backbone.interactions.3.interaction_units_12.branch1to2_injector.attn.sampling_offsets.bias", + "backbone.interactions.3.interaction_units_12.branch1to2_injector.attn.attention_weights.bias", + "backbone.interactions.3.interaction_units_12.branch1to2_injector.attn.value_proj.bias", + "backbone.interactions.3.interaction_units_12.branch1to2_injector.attn.output_proj.bias", + "backbone.interactions.3.interaction_units_12.branch1to2_injector.ffn.fc1.bias", + "backbone.interactions.3.interaction_units_12.branch1to2_injector.ffn.dwconv.dwconv.bias", + "backbone.interactions.3.interaction_units_12.branch1to2_injector.ffn.fc2.bias", + "backbone.interactions.3.interaction_units_12.branch1to2_injector.ffn_norm.weight", + "backbone.interactions.3.interaction_units_12.branch1to2_injector.ffn_norm.bias", + "backbone.interactions.3.interaction_units_23.branch2to1_proj.bias", + "backbone.interactions.3.interaction_units_23.branch2to1_injector.ca_gamma", + "backbone.interactions.3.interaction_units_23.branch2to1_injector.cffn_gamma", + "backbone.interactions.3.interaction_units_23.branch2to1_injector.query_norm.weight", + "backbone.interactions.3.interaction_units_23.branch2to1_injector.query_norm.bias", + "backbone.interactions.3.interaction_units_23.branch2to1_injector.feat_norm.weight", + "backbone.interactions.3.interaction_units_23.branch2to1_injector.feat_norm.bias", + "backbone.interactions.3.interaction_units_23.branch2to1_injector.attn.sampling_offsets.bias", + "backbone.interactions.3.interaction_units_23.branch2to1_injector.attn.attention_weights.bias", + "backbone.interactions.3.interaction_units_23.branch2to1_injector.attn.value_proj.bias", + "backbone.interactions.3.interaction_units_23.branch2to1_injector.attn.output_proj.bias", + "backbone.interactions.3.interaction_units_23.branch2to1_injector.ffn.fc1.bias", + "backbone.interactions.3.interaction_units_23.branch2to1_injector.ffn.dwconv.dwconv.bias", + "backbone.interactions.3.interaction_units_23.branch2to1_injector.ffn.fc2.bias", + "backbone.interactions.3.interaction_units_23.branch2to1_injector.ffn_norm.weight", + "backbone.interactions.3.interaction_units_23.branch2to1_injector.ffn_norm.bias", + "backbone.interactions.3.interaction_units_23.branch1to2_proj.bias", + "backbone.interactions.3.interaction_units_23.branch1to2_injector.ca_gamma", + "backbone.interactions.3.interaction_units_23.branch1to2_injector.cffn_gamma", + "backbone.interactions.3.interaction_units_23.branch1to2_injector.query_norm.weight", + "backbone.interactions.3.interaction_units_23.branch1to2_injector.query_norm.bias", + "backbone.interactions.3.interaction_units_23.branch1to2_injector.feat_norm.weight", + "backbone.interactions.3.interaction_units_23.branch1to2_injector.feat_norm.bias", + "backbone.interactions.3.interaction_units_23.branch1to2_injector.attn.sampling_offsets.bias", + "backbone.interactions.3.interaction_units_23.branch1to2_injector.attn.attention_weights.bias", + "backbone.interactions.3.interaction_units_23.branch1to2_injector.attn.value_proj.bias", + "backbone.interactions.3.interaction_units_23.branch1to2_injector.attn.output_proj.bias", + "backbone.interactions.3.interaction_units_23.branch1to2_injector.ffn.fc1.bias", + "backbone.interactions.3.interaction_units_23.branch1to2_injector.ffn.dwconv.dwconv.bias", + "backbone.interactions.3.interaction_units_23.branch1to2_injector.ffn.fc2.bias", + "backbone.interactions.3.interaction_units_23.branch1to2_injector.ffn_norm.weight", + "backbone.interactions.3.interaction_units_23.branch1to2_injector.ffn_norm.bias", + "backbone.interactions.4.interaction_units_12.branch2to1_proj.bias", + "backbone.interactions.4.interaction_units_12.branch2to1_injector.ca_gamma", + "backbone.interactions.4.interaction_units_12.branch2to1_injector.cffn_gamma", + "backbone.interactions.4.interaction_units_12.branch2to1_injector.query_norm.weight", + "backbone.interactions.4.interaction_units_12.branch2to1_injector.query_norm.bias", + "backbone.interactions.4.interaction_units_12.branch2to1_injector.feat_norm.weight", + "backbone.interactions.4.interaction_units_12.branch2to1_injector.feat_norm.bias", + "backbone.interactions.4.interaction_units_12.branch2to1_injector.attn.sampling_offsets.bias", + "backbone.interactions.4.interaction_units_12.branch2to1_injector.attn.attention_weights.bias", + "backbone.interactions.4.interaction_units_12.branch2to1_injector.attn.value_proj.bias", + "backbone.interactions.4.interaction_units_12.branch2to1_injector.attn.output_proj.bias", + "backbone.interactions.4.interaction_units_12.branch2to1_injector.ffn.fc1.bias", + "backbone.interactions.4.interaction_units_12.branch2to1_injector.ffn.dwconv.dwconv.bias", + "backbone.interactions.4.interaction_units_12.branch2to1_injector.ffn.fc2.bias", + "backbone.interactions.4.interaction_units_12.branch2to1_injector.ffn_norm.weight", + "backbone.interactions.4.interaction_units_12.branch2to1_injector.ffn_norm.bias", + "backbone.interactions.4.interaction_units_12.branch1to2_proj.bias", + "backbone.interactions.4.interaction_units_12.branch1to2_injector.ca_gamma", + "backbone.interactions.4.interaction_units_12.branch1to2_injector.cffn_gamma", + "backbone.interactions.4.interaction_units_12.branch1to2_injector.query_norm.weight", + "backbone.interactions.4.interaction_units_12.branch1to2_injector.query_norm.bias", + "backbone.interactions.4.interaction_units_12.branch1to2_injector.feat_norm.weight", + "backbone.interactions.4.interaction_units_12.branch1to2_injector.feat_norm.bias", + "backbone.interactions.4.interaction_units_12.branch1to2_injector.attn.sampling_offsets.bias", + "backbone.interactions.4.interaction_units_12.branch1to2_injector.attn.attention_weights.bias", + "backbone.interactions.4.interaction_units_12.branch1to2_injector.attn.value_proj.bias", + "backbone.interactions.4.interaction_units_12.branch1to2_injector.attn.output_proj.bias", + "backbone.interactions.4.interaction_units_12.branch1to2_injector.ffn.fc1.bias", + "backbone.interactions.4.interaction_units_12.branch1to2_injector.ffn.dwconv.dwconv.bias", + "backbone.interactions.4.interaction_units_12.branch1to2_injector.ffn.fc2.bias", + "backbone.interactions.4.interaction_units_12.branch1to2_injector.ffn_norm.weight", + "backbone.interactions.4.interaction_units_12.branch1to2_injector.ffn_norm.bias", + "backbone.interactions.4.interaction_units_23.branch2to1_proj.bias", + "backbone.interactions.4.interaction_units_23.branch2to1_injector.ca_gamma", + "backbone.interactions.4.interaction_units_23.branch2to1_injector.cffn_gamma", + "backbone.interactions.4.interaction_units_23.branch2to1_injector.query_norm.weight", + "backbone.interactions.4.interaction_units_23.branch2to1_injector.query_norm.bias", + "backbone.interactions.4.interaction_units_23.branch2to1_injector.feat_norm.weight", + "backbone.interactions.4.interaction_units_23.branch2to1_injector.feat_norm.bias", + "backbone.interactions.4.interaction_units_23.branch2to1_injector.attn.sampling_offsets.bias", + "backbone.interactions.4.interaction_units_23.branch2to1_injector.attn.attention_weights.bias", + "backbone.interactions.4.interaction_units_23.branch2to1_injector.attn.value_proj.bias", + "backbone.interactions.4.interaction_units_23.branch2to1_injector.attn.output_proj.bias", + "backbone.interactions.4.interaction_units_23.branch2to1_injector.ffn.fc1.bias", + "backbone.interactions.4.interaction_units_23.branch2to1_injector.ffn.dwconv.dwconv.bias", + "backbone.interactions.4.interaction_units_23.branch2to1_injector.ffn.fc2.bias", + "backbone.interactions.4.interaction_units_23.branch2to1_injector.ffn_norm.weight", + "backbone.interactions.4.interaction_units_23.branch2to1_injector.ffn_norm.bias", + "backbone.interactions.4.interaction_units_23.branch1to2_proj.bias", + "backbone.interactions.4.interaction_units_23.branch1to2_injector.ca_gamma", + "backbone.interactions.4.interaction_units_23.branch1to2_injector.cffn_gamma", + "backbone.interactions.4.interaction_units_23.branch1to2_injector.query_norm.weight", + "backbone.interactions.4.interaction_units_23.branch1to2_injector.query_norm.bias", + "backbone.interactions.4.interaction_units_23.branch1to2_injector.feat_norm.weight", + "backbone.interactions.4.interaction_units_23.branch1to2_injector.feat_norm.bias", + "backbone.interactions.4.interaction_units_23.branch1to2_injector.attn.sampling_offsets.bias", + "backbone.interactions.4.interaction_units_23.branch1to2_injector.attn.attention_weights.bias", + "backbone.interactions.4.interaction_units_23.branch1to2_injector.attn.value_proj.bias", + "backbone.interactions.4.interaction_units_23.branch1to2_injector.attn.output_proj.bias", + "backbone.interactions.4.interaction_units_23.branch1to2_injector.ffn.fc1.bias", + "backbone.interactions.4.interaction_units_23.branch1to2_injector.ffn.dwconv.dwconv.bias", + "backbone.interactions.4.interaction_units_23.branch1to2_injector.ffn.fc2.bias", + "backbone.interactions.4.interaction_units_23.branch1to2_injector.ffn_norm.weight", + "backbone.interactions.4.interaction_units_23.branch1to2_injector.ffn_norm.bias", + "backbone.interactions.5.interaction_units_12.branch2to1_proj.bias", + "backbone.interactions.5.interaction_units_12.branch2to1_injector.ca_gamma", + "backbone.interactions.5.interaction_units_12.branch2to1_injector.cffn_gamma", + "backbone.interactions.5.interaction_units_12.branch2to1_injector.query_norm.weight", + "backbone.interactions.5.interaction_units_12.branch2to1_injector.query_norm.bias", + "backbone.interactions.5.interaction_units_12.branch2to1_injector.feat_norm.weight", + "backbone.interactions.5.interaction_units_12.branch2to1_injector.feat_norm.bias", + "backbone.interactions.5.interaction_units_12.branch2to1_injector.attn.sampling_offsets.bias", + "backbone.interactions.5.interaction_units_12.branch2to1_injector.attn.attention_weights.bias", + "backbone.interactions.5.interaction_units_12.branch2to1_injector.attn.value_proj.bias", + "backbone.interactions.5.interaction_units_12.branch2to1_injector.attn.output_proj.bias", + "backbone.interactions.5.interaction_units_12.branch2to1_injector.ffn.fc1.bias", + "backbone.interactions.5.interaction_units_12.branch2to1_injector.ffn.dwconv.dwconv.bias", + "backbone.interactions.5.interaction_units_12.branch2to1_injector.ffn.fc2.bias", + "backbone.interactions.5.interaction_units_12.branch2to1_injector.ffn_norm.weight", + "backbone.interactions.5.interaction_units_12.branch2to1_injector.ffn_norm.bias", + "backbone.interactions.5.interaction_units_12.branch1to2_proj.bias", + "backbone.interactions.5.interaction_units_12.branch1to2_injector.ca_gamma", + "backbone.interactions.5.interaction_units_12.branch1to2_injector.cffn_gamma", + "backbone.interactions.5.interaction_units_12.branch1to2_injector.query_norm.weight", + "backbone.interactions.5.interaction_units_12.branch1to2_injector.query_norm.bias", + "backbone.interactions.5.interaction_units_12.branch1to2_injector.feat_norm.weight", + "backbone.interactions.5.interaction_units_12.branch1to2_injector.feat_norm.bias", + "backbone.interactions.5.interaction_units_12.branch1to2_injector.attn.sampling_offsets.bias", + "backbone.interactions.5.interaction_units_12.branch1to2_injector.attn.attention_weights.bias", + "backbone.interactions.5.interaction_units_12.branch1to2_injector.attn.value_proj.bias", + "backbone.interactions.5.interaction_units_12.branch1to2_injector.attn.output_proj.bias", + "backbone.interactions.5.interaction_units_12.branch1to2_injector.ffn.fc1.bias", + "backbone.interactions.5.interaction_units_12.branch1to2_injector.ffn.dwconv.dwconv.bias", + "backbone.interactions.5.interaction_units_12.branch1to2_injector.ffn.fc2.bias", + "backbone.interactions.5.interaction_units_12.branch1to2_injector.ffn_norm.weight", + "backbone.interactions.5.interaction_units_12.branch1to2_injector.ffn_norm.bias", + "backbone.interactions.5.interaction_units_23.branch2to1_proj.bias", + "backbone.interactions.5.interaction_units_23.branch2to1_injector.ca_gamma", + "backbone.interactions.5.interaction_units_23.branch2to1_injector.cffn_gamma", + "backbone.interactions.5.interaction_units_23.branch2to1_injector.query_norm.weight", + "backbone.interactions.5.interaction_units_23.branch2to1_injector.query_norm.bias", + "backbone.interactions.5.interaction_units_23.branch2to1_injector.feat_norm.weight", + "backbone.interactions.5.interaction_units_23.branch2to1_injector.feat_norm.bias", + "backbone.interactions.5.interaction_units_23.branch2to1_injector.attn.sampling_offsets.bias", + "backbone.interactions.5.interaction_units_23.branch2to1_injector.attn.attention_weights.bias", + "backbone.interactions.5.interaction_units_23.branch2to1_injector.attn.value_proj.bias", + "backbone.interactions.5.interaction_units_23.branch2to1_injector.attn.output_proj.bias", + "backbone.interactions.5.interaction_units_23.branch2to1_injector.ffn.fc1.bias", + "backbone.interactions.5.interaction_units_23.branch2to1_injector.ffn.dwconv.dwconv.bias", + "backbone.interactions.5.interaction_units_23.branch2to1_injector.ffn.fc2.bias", + "backbone.interactions.5.interaction_units_23.branch2to1_injector.ffn_norm.weight", + "backbone.interactions.5.interaction_units_23.branch2to1_injector.ffn_norm.bias", + "backbone.interactions.5.interaction_units_23.branch1to2_proj.bias", + "backbone.interactions.5.interaction_units_23.branch1to2_injector.ca_gamma", + "backbone.interactions.5.interaction_units_23.branch1to2_injector.cffn_gamma", + "backbone.interactions.5.interaction_units_23.branch1to2_injector.query_norm.weight", + "backbone.interactions.5.interaction_units_23.branch1to2_injector.query_norm.bias", + "backbone.interactions.5.interaction_units_23.branch1to2_injector.feat_norm.weight", + "backbone.interactions.5.interaction_units_23.branch1to2_injector.feat_norm.bias", + "backbone.interactions.5.interaction_units_23.branch1to2_injector.attn.sampling_offsets.bias", + "backbone.interactions.5.interaction_units_23.branch1to2_injector.attn.attention_weights.bias", + "backbone.interactions.5.interaction_units_23.branch1to2_injector.attn.value_proj.bias", + "backbone.interactions.5.interaction_units_23.branch1to2_injector.attn.output_proj.bias", + "backbone.interactions.5.interaction_units_23.branch1to2_injector.ffn.fc1.bias", + "backbone.interactions.5.interaction_units_23.branch1to2_injector.ffn.dwconv.dwconv.bias", + "backbone.interactions.5.interaction_units_23.branch1to2_injector.ffn.fc2.bias", + "backbone.interactions.5.interaction_units_23.branch1to2_injector.ffn_norm.weight", + "backbone.interactions.5.interaction_units_23.branch1to2_injector.ffn_norm.bias", + "backbone.interactions.6.interaction_units_12.branch2to1_proj.bias", + "backbone.interactions.6.interaction_units_12.branch2to1_injector.ca_gamma", + "backbone.interactions.6.interaction_units_12.branch2to1_injector.cffn_gamma", + "backbone.interactions.6.interaction_units_12.branch2to1_injector.query_norm.weight", + "backbone.interactions.6.interaction_units_12.branch2to1_injector.query_norm.bias", + "backbone.interactions.6.interaction_units_12.branch2to1_injector.feat_norm.weight", + "backbone.interactions.6.interaction_units_12.branch2to1_injector.feat_norm.bias", + "backbone.interactions.6.interaction_units_12.branch2to1_injector.attn.sampling_offsets.bias", + "backbone.interactions.6.interaction_units_12.branch2to1_injector.attn.attention_weights.bias", + "backbone.interactions.6.interaction_units_12.branch2to1_injector.attn.value_proj.bias", + "backbone.interactions.6.interaction_units_12.branch2to1_injector.attn.output_proj.bias", + "backbone.interactions.6.interaction_units_12.branch2to1_injector.ffn.fc1.bias", + "backbone.interactions.6.interaction_units_12.branch2to1_injector.ffn.dwconv.dwconv.bias", + "backbone.interactions.6.interaction_units_12.branch2to1_injector.ffn.fc2.bias", + "backbone.interactions.6.interaction_units_12.branch2to1_injector.ffn_norm.weight", + "backbone.interactions.6.interaction_units_12.branch2to1_injector.ffn_norm.bias", + "backbone.interactions.6.interaction_units_12.branch1to2_proj.bias", + "backbone.interactions.6.interaction_units_12.branch1to2_injector.ca_gamma", + "backbone.interactions.6.interaction_units_12.branch1to2_injector.cffn_gamma", + "backbone.interactions.6.interaction_units_12.branch1to2_injector.query_norm.weight", + "backbone.interactions.6.interaction_units_12.branch1to2_injector.query_norm.bias", + "backbone.interactions.6.interaction_units_12.branch1to2_injector.feat_norm.weight", + "backbone.interactions.6.interaction_units_12.branch1to2_injector.feat_norm.bias", + "backbone.interactions.6.interaction_units_12.branch1to2_injector.attn.sampling_offsets.bias", + "backbone.interactions.6.interaction_units_12.branch1to2_injector.attn.attention_weights.bias", + "backbone.interactions.6.interaction_units_12.branch1to2_injector.attn.value_proj.bias", + "backbone.interactions.6.interaction_units_12.branch1to2_injector.attn.output_proj.bias", + "backbone.interactions.6.interaction_units_12.branch1to2_injector.ffn.fc1.bias", + "backbone.interactions.6.interaction_units_12.branch1to2_injector.ffn.dwconv.dwconv.bias", + "backbone.interactions.6.interaction_units_12.branch1to2_injector.ffn.fc2.bias", + "backbone.interactions.6.interaction_units_12.branch1to2_injector.ffn_norm.weight", + "backbone.interactions.6.interaction_units_12.branch1to2_injector.ffn_norm.bias", + "backbone.interactions.6.interaction_units_23.branch2to1_proj.bias", + "backbone.interactions.6.interaction_units_23.branch2to1_injector.ca_gamma", + "backbone.interactions.6.interaction_units_23.branch2to1_injector.cffn_gamma", + "backbone.interactions.6.interaction_units_23.branch2to1_injector.query_norm.weight", + "backbone.interactions.6.interaction_units_23.branch2to1_injector.query_norm.bias", + "backbone.interactions.6.interaction_units_23.branch2to1_injector.feat_norm.weight", + "backbone.interactions.6.interaction_units_23.branch2to1_injector.feat_norm.bias", + "backbone.interactions.6.interaction_units_23.branch2to1_injector.attn.sampling_offsets.bias", + "backbone.interactions.6.interaction_units_23.branch2to1_injector.attn.attention_weights.bias", + "backbone.interactions.6.interaction_units_23.branch2to1_injector.attn.value_proj.bias", + "backbone.interactions.6.interaction_units_23.branch2to1_injector.attn.output_proj.bias", + "backbone.interactions.6.interaction_units_23.branch2to1_injector.ffn.fc1.bias", + "backbone.interactions.6.interaction_units_23.branch2to1_injector.ffn.dwconv.dwconv.bias", + "backbone.interactions.6.interaction_units_23.branch2to1_injector.ffn.fc2.bias", + "backbone.interactions.6.interaction_units_23.branch2to1_injector.ffn_norm.weight", + "backbone.interactions.6.interaction_units_23.branch2to1_injector.ffn_norm.bias", + "backbone.interactions.6.interaction_units_23.branch1to2_proj.bias", + "backbone.interactions.6.interaction_units_23.branch1to2_injector.ca_gamma", + "backbone.interactions.6.interaction_units_23.branch1to2_injector.cffn_gamma", + "backbone.interactions.6.interaction_units_23.branch1to2_injector.query_norm.weight", + "backbone.interactions.6.interaction_units_23.branch1to2_injector.query_norm.bias", + "backbone.interactions.6.interaction_units_23.branch1to2_injector.feat_norm.weight", + "backbone.interactions.6.interaction_units_23.branch1to2_injector.feat_norm.bias", + "backbone.interactions.6.interaction_units_23.branch1to2_injector.attn.sampling_offsets.bias", + "backbone.interactions.6.interaction_units_23.branch1to2_injector.attn.attention_weights.bias", + "backbone.interactions.6.interaction_units_23.branch1to2_injector.attn.value_proj.bias", + "backbone.interactions.6.interaction_units_23.branch1to2_injector.attn.output_proj.bias", + "backbone.interactions.6.interaction_units_23.branch1to2_injector.ffn.fc1.bias", + "backbone.interactions.6.interaction_units_23.branch1to2_injector.ffn.dwconv.dwconv.bias", + "backbone.interactions.6.interaction_units_23.branch1to2_injector.ffn.fc2.bias", + "backbone.interactions.6.interaction_units_23.branch1to2_injector.ffn_norm.weight", + "backbone.interactions.6.interaction_units_23.branch1to2_injector.ffn_norm.bias", + "backbone.interactions.7.interaction_units_12.branch2to1_proj.bias", + "backbone.interactions.7.interaction_units_12.branch2to1_injector.ca_gamma", + "backbone.interactions.7.interaction_units_12.branch2to1_injector.cffn_gamma", + "backbone.interactions.7.interaction_units_12.branch2to1_injector.query_norm.weight", + "backbone.interactions.7.interaction_units_12.branch2to1_injector.query_norm.bias", + "backbone.interactions.7.interaction_units_12.branch2to1_injector.feat_norm.weight", + "backbone.interactions.7.interaction_units_12.branch2to1_injector.feat_norm.bias", + "backbone.interactions.7.interaction_units_12.branch2to1_injector.attn.sampling_offsets.bias", + "backbone.interactions.7.interaction_units_12.branch2to1_injector.attn.attention_weights.bias", + "backbone.interactions.7.interaction_units_12.branch2to1_injector.attn.value_proj.bias", + "backbone.interactions.7.interaction_units_12.branch2to1_injector.attn.output_proj.bias", + "backbone.interactions.7.interaction_units_12.branch2to1_injector.ffn.fc1.bias", + "backbone.interactions.7.interaction_units_12.branch2to1_injector.ffn.dwconv.dwconv.bias", + "backbone.interactions.7.interaction_units_12.branch2to1_injector.ffn.fc2.bias", + "backbone.interactions.7.interaction_units_12.branch2to1_injector.ffn_norm.weight", + "backbone.interactions.7.interaction_units_12.branch2to1_injector.ffn_norm.bias", + "backbone.interactions.7.interaction_units_12.branch1to2_proj.bias", + "backbone.interactions.7.interaction_units_12.branch1to2_injector.ca_gamma", + "backbone.interactions.7.interaction_units_12.branch1to2_injector.cffn_gamma", + "backbone.interactions.7.interaction_units_12.branch1to2_injector.query_norm.weight", + "backbone.interactions.7.interaction_units_12.branch1to2_injector.query_norm.bias", + "backbone.interactions.7.interaction_units_12.branch1to2_injector.feat_norm.weight", + "backbone.interactions.7.interaction_units_12.branch1to2_injector.feat_norm.bias", + "backbone.interactions.7.interaction_units_12.branch1to2_injector.attn.sampling_offsets.bias", + "backbone.interactions.7.interaction_units_12.branch1to2_injector.attn.attention_weights.bias", + "backbone.interactions.7.interaction_units_12.branch1to2_injector.attn.value_proj.bias", + "backbone.interactions.7.interaction_units_12.branch1to2_injector.attn.output_proj.bias", + "backbone.interactions.7.interaction_units_12.branch1to2_injector.ffn.fc1.bias", + "backbone.interactions.7.interaction_units_12.branch1to2_injector.ffn.dwconv.dwconv.bias", + "backbone.interactions.7.interaction_units_12.branch1to2_injector.ffn.fc2.bias", + "backbone.interactions.7.interaction_units_12.branch1to2_injector.ffn_norm.weight", + "backbone.interactions.7.interaction_units_12.branch1to2_injector.ffn_norm.bias", + "backbone.interactions.7.interaction_units_23.branch2to1_proj.bias", + "backbone.interactions.7.interaction_units_23.branch2to1_injector.ca_gamma", + "backbone.interactions.7.interaction_units_23.branch2to1_injector.cffn_gamma", + "backbone.interactions.7.interaction_units_23.branch2to1_injector.query_norm.weight", + "backbone.interactions.7.interaction_units_23.branch2to1_injector.query_norm.bias", + "backbone.interactions.7.interaction_units_23.branch2to1_injector.feat_norm.weight", + "backbone.interactions.7.interaction_units_23.branch2to1_injector.feat_norm.bias", + "backbone.interactions.7.interaction_units_23.branch2to1_injector.attn.sampling_offsets.bias", + "backbone.interactions.7.interaction_units_23.branch2to1_injector.attn.attention_weights.bias", + "backbone.interactions.7.interaction_units_23.branch2to1_injector.attn.value_proj.bias", + "backbone.interactions.7.interaction_units_23.branch2to1_injector.attn.output_proj.bias", + "backbone.interactions.7.interaction_units_23.branch2to1_injector.ffn.fc1.bias", + "backbone.interactions.7.interaction_units_23.branch2to1_injector.ffn.dwconv.dwconv.bias", + "backbone.interactions.7.interaction_units_23.branch2to1_injector.ffn.fc2.bias", + "backbone.interactions.7.interaction_units_23.branch2to1_injector.ffn_norm.weight", + "backbone.interactions.7.interaction_units_23.branch2to1_injector.ffn_norm.bias", + "backbone.interactions.7.interaction_units_23.branch1to2_proj.bias", + "backbone.interactions.7.interaction_units_23.branch1to2_injector.ca_gamma", + "backbone.interactions.7.interaction_units_23.branch1to2_injector.cffn_gamma", + "backbone.interactions.7.interaction_units_23.branch1to2_injector.query_norm.weight", + "backbone.interactions.7.interaction_units_23.branch1to2_injector.query_norm.bias", + "backbone.interactions.7.interaction_units_23.branch1to2_injector.feat_norm.weight", + "backbone.interactions.7.interaction_units_23.branch1to2_injector.feat_norm.bias", + "backbone.interactions.7.interaction_units_23.branch1to2_injector.attn.sampling_offsets.bias", + "backbone.interactions.7.interaction_units_23.branch1to2_injector.attn.attention_weights.bias", + "backbone.interactions.7.interaction_units_23.branch1to2_injector.attn.value_proj.bias", + "backbone.interactions.7.interaction_units_23.branch1to2_injector.attn.output_proj.bias", + "backbone.interactions.7.interaction_units_23.branch1to2_injector.ffn.fc1.bias", + "backbone.interactions.7.interaction_units_23.branch1to2_injector.ffn.dwconv.dwconv.bias", + "backbone.interactions.7.interaction_units_23.branch1to2_injector.ffn.fc2.bias", + "backbone.interactions.7.interaction_units_23.branch1to2_injector.ffn_norm.weight", + "backbone.interactions.7.interaction_units_23.branch1to2_injector.ffn_norm.bias", + "backbone.interactions.8.interaction_units_12.branch2to1_proj.bias", + "backbone.interactions.8.interaction_units_12.branch2to1_injector.ca_gamma", + "backbone.interactions.8.interaction_units_12.branch2to1_injector.cffn_gamma", + "backbone.interactions.8.interaction_units_12.branch2to1_injector.query_norm.weight", + "backbone.interactions.8.interaction_units_12.branch2to1_injector.query_norm.bias", + "backbone.interactions.8.interaction_units_12.branch2to1_injector.feat_norm.weight", + "backbone.interactions.8.interaction_units_12.branch2to1_injector.feat_norm.bias", + "backbone.interactions.8.interaction_units_12.branch2to1_injector.attn.sampling_offsets.bias", + "backbone.interactions.8.interaction_units_12.branch2to1_injector.attn.attention_weights.bias", + "backbone.interactions.8.interaction_units_12.branch2to1_injector.attn.value_proj.bias", + "backbone.interactions.8.interaction_units_12.branch2to1_injector.attn.output_proj.bias", + "backbone.interactions.8.interaction_units_12.branch2to1_injector.ffn.fc1.bias", + "backbone.interactions.8.interaction_units_12.branch2to1_injector.ffn.dwconv.dwconv.bias", + "backbone.interactions.8.interaction_units_12.branch2to1_injector.ffn.fc2.bias", + "backbone.interactions.8.interaction_units_12.branch2to1_injector.ffn_norm.weight", + "backbone.interactions.8.interaction_units_12.branch2to1_injector.ffn_norm.bias", + "backbone.interactions.8.interaction_units_12.branch1to2_proj.bias", + "backbone.interactions.8.interaction_units_12.branch1to2_injector.ca_gamma", + "backbone.interactions.8.interaction_units_12.branch1to2_injector.cffn_gamma", + "backbone.interactions.8.interaction_units_12.branch1to2_injector.query_norm.weight", + "backbone.interactions.8.interaction_units_12.branch1to2_injector.query_norm.bias", + "backbone.interactions.8.interaction_units_12.branch1to2_injector.feat_norm.weight", + "backbone.interactions.8.interaction_units_12.branch1to2_injector.feat_norm.bias", + "backbone.interactions.8.interaction_units_12.branch1to2_injector.attn.sampling_offsets.bias", + "backbone.interactions.8.interaction_units_12.branch1to2_injector.attn.attention_weights.bias", + "backbone.interactions.8.interaction_units_12.branch1to2_injector.attn.value_proj.bias", + "backbone.interactions.8.interaction_units_12.branch1to2_injector.attn.output_proj.bias", + "backbone.interactions.8.interaction_units_12.branch1to2_injector.ffn.fc1.bias", + "backbone.interactions.8.interaction_units_12.branch1to2_injector.ffn.dwconv.dwconv.bias", + "backbone.interactions.8.interaction_units_12.branch1to2_injector.ffn.fc2.bias", + "backbone.interactions.8.interaction_units_12.branch1to2_injector.ffn_norm.weight", + "backbone.interactions.8.interaction_units_12.branch1to2_injector.ffn_norm.bias", + "backbone.interactions.8.interaction_units_23.branch2to1_proj.bias", + "backbone.interactions.8.interaction_units_23.branch2to1_injector.ca_gamma", + "backbone.interactions.8.interaction_units_23.branch2to1_injector.cffn_gamma", + "backbone.interactions.8.interaction_units_23.branch2to1_injector.query_norm.weight", + "backbone.interactions.8.interaction_units_23.branch2to1_injector.query_norm.bias", + "backbone.interactions.8.interaction_units_23.branch2to1_injector.feat_norm.weight", + "backbone.interactions.8.interaction_units_23.branch2to1_injector.feat_norm.bias", + "backbone.interactions.8.interaction_units_23.branch2to1_injector.attn.sampling_offsets.bias", + "backbone.interactions.8.interaction_units_23.branch2to1_injector.attn.attention_weights.bias", + "backbone.interactions.8.interaction_units_23.branch2to1_injector.attn.value_proj.bias", + "backbone.interactions.8.interaction_units_23.branch2to1_injector.attn.output_proj.bias", + "backbone.interactions.8.interaction_units_23.branch2to1_injector.ffn.fc1.bias", + "backbone.interactions.8.interaction_units_23.branch2to1_injector.ffn.dwconv.dwconv.bias", + "backbone.interactions.8.interaction_units_23.branch2to1_injector.ffn.fc2.bias", + "backbone.interactions.8.interaction_units_23.branch2to1_injector.ffn_norm.weight", + "backbone.interactions.8.interaction_units_23.branch2to1_injector.ffn_norm.bias", + "backbone.interactions.8.interaction_units_23.branch1to2_proj.bias", + "backbone.interactions.8.interaction_units_23.branch1to2_injector.ca_gamma", + "backbone.interactions.8.interaction_units_23.branch1to2_injector.cffn_gamma", + "backbone.interactions.8.interaction_units_23.branch1to2_injector.query_norm.weight", + "backbone.interactions.8.interaction_units_23.branch1to2_injector.query_norm.bias", + "backbone.interactions.8.interaction_units_23.branch1to2_injector.feat_norm.weight", + "backbone.interactions.8.interaction_units_23.branch1to2_injector.feat_norm.bias", + "backbone.interactions.8.interaction_units_23.branch1to2_injector.attn.sampling_offsets.bias", + "backbone.interactions.8.interaction_units_23.branch1to2_injector.attn.attention_weights.bias", + "backbone.interactions.8.interaction_units_23.branch1to2_injector.attn.value_proj.bias", + "backbone.interactions.8.interaction_units_23.branch1to2_injector.attn.output_proj.bias", + "backbone.interactions.8.interaction_units_23.branch1to2_injector.ffn.fc1.bias", + "backbone.interactions.8.interaction_units_23.branch1to2_injector.ffn.dwconv.dwconv.bias", + "backbone.interactions.8.interaction_units_23.branch1to2_injector.ffn.fc2.bias", + "backbone.interactions.8.interaction_units_23.branch1to2_injector.ffn_norm.weight", + "backbone.interactions.8.interaction_units_23.branch1to2_injector.ffn_norm.bias", + "backbone.interactions.9.interaction_units_12.branch2to1_proj.bias", + "backbone.interactions.9.interaction_units_12.branch2to1_injector.ca_gamma", + "backbone.interactions.9.interaction_units_12.branch2to1_injector.cffn_gamma", + "backbone.interactions.9.interaction_units_12.branch2to1_injector.query_norm.weight", + "backbone.interactions.9.interaction_units_12.branch2to1_injector.query_norm.bias", + "backbone.interactions.9.interaction_units_12.branch2to1_injector.feat_norm.weight", + "backbone.interactions.9.interaction_units_12.branch2to1_injector.feat_norm.bias", + "backbone.interactions.9.interaction_units_12.branch2to1_injector.attn.sampling_offsets.bias", + "backbone.interactions.9.interaction_units_12.branch2to1_injector.attn.attention_weights.bias", + "backbone.interactions.9.interaction_units_12.branch2to1_injector.attn.value_proj.bias", + "backbone.interactions.9.interaction_units_12.branch2to1_injector.attn.output_proj.bias", + "backbone.interactions.9.interaction_units_12.branch2to1_injector.ffn.fc1.bias", + "backbone.interactions.9.interaction_units_12.branch2to1_injector.ffn.dwconv.dwconv.bias", + "backbone.interactions.9.interaction_units_12.branch2to1_injector.ffn.fc2.bias", + "backbone.interactions.9.interaction_units_12.branch2to1_injector.ffn_norm.weight", + "backbone.interactions.9.interaction_units_12.branch2to1_injector.ffn_norm.bias", + "backbone.interactions.9.interaction_units_12.branch1to2_proj.bias", + "backbone.interactions.9.interaction_units_12.branch1to2_injector.ca_gamma", + "backbone.interactions.9.interaction_units_12.branch1to2_injector.cffn_gamma", + "backbone.interactions.9.interaction_units_12.branch1to2_injector.query_norm.weight", + "backbone.interactions.9.interaction_units_12.branch1to2_injector.query_norm.bias", + "backbone.interactions.9.interaction_units_12.branch1to2_injector.feat_norm.weight", + "backbone.interactions.9.interaction_units_12.branch1to2_injector.feat_norm.bias", + "backbone.interactions.9.interaction_units_12.branch1to2_injector.attn.sampling_offsets.bias", + "backbone.interactions.9.interaction_units_12.branch1to2_injector.attn.attention_weights.bias", + "backbone.interactions.9.interaction_units_12.branch1to2_injector.attn.value_proj.bias", + "backbone.interactions.9.interaction_units_12.branch1to2_injector.attn.output_proj.bias", + "backbone.interactions.9.interaction_units_12.branch1to2_injector.ffn.fc1.bias", + "backbone.interactions.9.interaction_units_12.branch1to2_injector.ffn.dwconv.dwconv.bias", + "backbone.interactions.9.interaction_units_12.branch1to2_injector.ffn.fc2.bias", + "backbone.interactions.9.interaction_units_12.branch1to2_injector.ffn_norm.weight", + "backbone.interactions.9.interaction_units_12.branch1to2_injector.ffn_norm.bias", + "backbone.interactions.9.interaction_units_23.branch2to1_proj.bias", + "backbone.interactions.9.interaction_units_23.branch2to1_injector.ca_gamma", + "backbone.interactions.9.interaction_units_23.branch2to1_injector.cffn_gamma", + "backbone.interactions.9.interaction_units_23.branch2to1_injector.query_norm.weight", + "backbone.interactions.9.interaction_units_23.branch2to1_injector.query_norm.bias", + "backbone.interactions.9.interaction_units_23.branch2to1_injector.feat_norm.weight", + "backbone.interactions.9.interaction_units_23.branch2to1_injector.feat_norm.bias", + "backbone.interactions.9.interaction_units_23.branch2to1_injector.attn.sampling_offsets.bias", + "backbone.interactions.9.interaction_units_23.branch2to1_injector.attn.attention_weights.bias", + "backbone.interactions.9.interaction_units_23.branch2to1_injector.attn.value_proj.bias", + "backbone.interactions.9.interaction_units_23.branch2to1_injector.attn.output_proj.bias", + "backbone.interactions.9.interaction_units_23.branch2to1_injector.ffn.fc1.bias", + "backbone.interactions.9.interaction_units_23.branch2to1_injector.ffn.dwconv.dwconv.bias", + "backbone.interactions.9.interaction_units_23.branch2to1_injector.ffn.fc2.bias", + "backbone.interactions.9.interaction_units_23.branch2to1_injector.ffn_norm.weight", + "backbone.interactions.9.interaction_units_23.branch2to1_injector.ffn_norm.bias", + "backbone.interactions.9.interaction_units_23.branch1to2_proj.bias", + "backbone.interactions.9.interaction_units_23.branch1to2_injector.ca_gamma", + "backbone.interactions.9.interaction_units_23.branch1to2_injector.cffn_gamma", + "backbone.interactions.9.interaction_units_23.branch1to2_injector.query_norm.weight", + "backbone.interactions.9.interaction_units_23.branch1to2_injector.query_norm.bias", + "backbone.interactions.9.interaction_units_23.branch1to2_injector.feat_norm.weight", + "backbone.interactions.9.interaction_units_23.branch1to2_injector.feat_norm.bias", + "backbone.interactions.9.interaction_units_23.branch1to2_injector.attn.sampling_offsets.bias", + "backbone.interactions.9.interaction_units_23.branch1to2_injector.attn.attention_weights.bias", + "backbone.interactions.9.interaction_units_23.branch1to2_injector.attn.value_proj.bias", + "backbone.interactions.9.interaction_units_23.branch1to2_injector.attn.output_proj.bias", + "backbone.interactions.9.interaction_units_23.branch1to2_injector.ffn.fc1.bias", + "backbone.interactions.9.interaction_units_23.branch1to2_injector.ffn.dwconv.dwconv.bias", + "backbone.interactions.9.interaction_units_23.branch1to2_injector.ffn.fc2.bias", + "backbone.interactions.9.interaction_units_23.branch1to2_injector.ffn_norm.weight", + "backbone.interactions.9.interaction_units_23.branch1to2_injector.ffn_norm.bias", + "backbone.interactions.10.interaction_units_12.branch2to1_proj.bias", + "backbone.interactions.10.interaction_units_12.branch2to1_injector.ca_gamma", + "backbone.interactions.10.interaction_units_12.branch2to1_injector.cffn_gamma", + "backbone.interactions.10.interaction_units_12.branch2to1_injector.query_norm.weight", + "backbone.interactions.10.interaction_units_12.branch2to1_injector.query_norm.bias", + "backbone.interactions.10.interaction_units_12.branch2to1_injector.feat_norm.weight", + "backbone.interactions.10.interaction_units_12.branch2to1_injector.feat_norm.bias", + "backbone.interactions.10.interaction_units_12.branch2to1_injector.attn.sampling_offsets.bias", + "backbone.interactions.10.interaction_units_12.branch2to1_injector.attn.attention_weights.bias", + "backbone.interactions.10.interaction_units_12.branch2to1_injector.attn.value_proj.bias", + "backbone.interactions.10.interaction_units_12.branch2to1_injector.attn.output_proj.bias", + "backbone.interactions.10.interaction_units_12.branch2to1_injector.ffn.fc1.bias", + "backbone.interactions.10.interaction_units_12.branch2to1_injector.ffn.dwconv.dwconv.bias", + "backbone.interactions.10.interaction_units_12.branch2to1_injector.ffn.fc2.bias", + "backbone.interactions.10.interaction_units_12.branch2to1_injector.ffn_norm.weight", + "backbone.interactions.10.interaction_units_12.branch2to1_injector.ffn_norm.bias", + "backbone.interactions.10.interaction_units_12.branch1to2_proj.bias", + "backbone.interactions.10.interaction_units_12.branch1to2_injector.ca_gamma", + "backbone.interactions.10.interaction_units_12.branch1to2_injector.cffn_gamma", + "backbone.interactions.10.interaction_units_12.branch1to2_injector.query_norm.weight", + "backbone.interactions.10.interaction_units_12.branch1to2_injector.query_norm.bias", + "backbone.interactions.10.interaction_units_12.branch1to2_injector.feat_norm.weight", + "backbone.interactions.10.interaction_units_12.branch1to2_injector.feat_norm.bias", + "backbone.interactions.10.interaction_units_12.branch1to2_injector.attn.sampling_offsets.bias", + "backbone.interactions.10.interaction_units_12.branch1to2_injector.attn.attention_weights.bias", + "backbone.interactions.10.interaction_units_12.branch1to2_injector.attn.value_proj.bias", + "backbone.interactions.10.interaction_units_12.branch1to2_injector.attn.output_proj.bias", + "backbone.interactions.10.interaction_units_12.branch1to2_injector.ffn.fc1.bias", + "backbone.interactions.10.interaction_units_12.branch1to2_injector.ffn.dwconv.dwconv.bias", + "backbone.interactions.10.interaction_units_12.branch1to2_injector.ffn.fc2.bias", + "backbone.interactions.10.interaction_units_12.branch1to2_injector.ffn_norm.weight", + "backbone.interactions.10.interaction_units_12.branch1to2_injector.ffn_norm.bias", + "backbone.interactions.10.interaction_units_23.branch2to1_proj.bias", + "backbone.interactions.10.interaction_units_23.branch2to1_injector.ca_gamma", + "backbone.interactions.10.interaction_units_23.branch2to1_injector.cffn_gamma", + "backbone.interactions.10.interaction_units_23.branch2to1_injector.query_norm.weight", + "backbone.interactions.10.interaction_units_23.branch2to1_injector.query_norm.bias", + "backbone.interactions.10.interaction_units_23.branch2to1_injector.feat_norm.weight", + "backbone.interactions.10.interaction_units_23.branch2to1_injector.feat_norm.bias", + "backbone.interactions.10.interaction_units_23.branch2to1_injector.attn.sampling_offsets.bias", + "backbone.interactions.10.interaction_units_23.branch2to1_injector.attn.attention_weights.bias", + "backbone.interactions.10.interaction_units_23.branch2to1_injector.attn.value_proj.bias", + "backbone.interactions.10.interaction_units_23.branch2to1_injector.attn.output_proj.bias", + "backbone.interactions.10.interaction_units_23.branch2to1_injector.ffn.fc1.bias", + "backbone.interactions.10.interaction_units_23.branch2to1_injector.ffn.dwconv.dwconv.bias", + "backbone.interactions.10.interaction_units_23.branch2to1_injector.ffn.fc2.bias", + "backbone.interactions.10.interaction_units_23.branch2to1_injector.ffn_norm.weight", + "backbone.interactions.10.interaction_units_23.branch2to1_injector.ffn_norm.bias", + "backbone.interactions.10.interaction_units_23.branch1to2_proj.bias", + "backbone.interactions.10.interaction_units_23.branch1to2_injector.ca_gamma", + "backbone.interactions.10.interaction_units_23.branch1to2_injector.cffn_gamma", + "backbone.interactions.10.interaction_units_23.branch1to2_injector.query_norm.weight", + "backbone.interactions.10.interaction_units_23.branch1to2_injector.query_norm.bias", + "backbone.interactions.10.interaction_units_23.branch1to2_injector.feat_norm.weight", + "backbone.interactions.10.interaction_units_23.branch1to2_injector.feat_norm.bias", + "backbone.interactions.10.interaction_units_23.branch1to2_injector.attn.sampling_offsets.bias", + "backbone.interactions.10.interaction_units_23.branch1to2_injector.attn.attention_weights.bias", + "backbone.interactions.10.interaction_units_23.branch1to2_injector.attn.value_proj.bias", + "backbone.interactions.10.interaction_units_23.branch1to2_injector.attn.output_proj.bias", + "backbone.interactions.10.interaction_units_23.branch1to2_injector.ffn.fc1.bias", + "backbone.interactions.10.interaction_units_23.branch1to2_injector.ffn.dwconv.dwconv.bias", + "backbone.interactions.10.interaction_units_23.branch1to2_injector.ffn.fc2.bias", + "backbone.interactions.10.interaction_units_23.branch1to2_injector.ffn_norm.weight", + "backbone.interactions.10.interaction_units_23.branch1to2_injector.ffn_norm.bias", + "backbone.interactions.11.interaction_units_12.branch2to1_proj.bias", + "backbone.interactions.11.interaction_units_12.branch2to1_injector.ca_gamma", + "backbone.interactions.11.interaction_units_12.branch2to1_injector.cffn_gamma", + "backbone.interactions.11.interaction_units_12.branch2to1_injector.query_norm.weight", + "backbone.interactions.11.interaction_units_12.branch2to1_injector.query_norm.bias", + "backbone.interactions.11.interaction_units_12.branch2to1_injector.feat_norm.weight", + "backbone.interactions.11.interaction_units_12.branch2to1_injector.feat_norm.bias", + "backbone.interactions.11.interaction_units_12.branch2to1_injector.attn.sampling_offsets.bias", + "backbone.interactions.11.interaction_units_12.branch2to1_injector.attn.attention_weights.bias", + "backbone.interactions.11.interaction_units_12.branch2to1_injector.attn.value_proj.bias", + "backbone.interactions.11.interaction_units_12.branch2to1_injector.attn.output_proj.bias", + "backbone.interactions.11.interaction_units_12.branch2to1_injector.ffn.fc1.bias", + "backbone.interactions.11.interaction_units_12.branch2to1_injector.ffn.dwconv.dwconv.bias", + "backbone.interactions.11.interaction_units_12.branch2to1_injector.ffn.fc2.bias", + "backbone.interactions.11.interaction_units_12.branch2to1_injector.ffn_norm.weight", + "backbone.interactions.11.interaction_units_12.branch2to1_injector.ffn_norm.bias", + "backbone.interactions.11.interaction_units_12.branch1to2_proj.bias", + "backbone.interactions.11.interaction_units_12.branch1to2_injector.ca_gamma", + "backbone.interactions.11.interaction_units_12.branch1to2_injector.cffn_gamma", + "backbone.interactions.11.interaction_units_12.branch1to2_injector.query_norm.weight", + "backbone.interactions.11.interaction_units_12.branch1to2_injector.query_norm.bias", + "backbone.interactions.11.interaction_units_12.branch1to2_injector.feat_norm.weight", + "backbone.interactions.11.interaction_units_12.branch1to2_injector.feat_norm.bias", + "backbone.interactions.11.interaction_units_12.branch1to2_injector.attn.sampling_offsets.bias", + "backbone.interactions.11.interaction_units_12.branch1to2_injector.attn.attention_weights.bias", + "backbone.interactions.11.interaction_units_12.branch1to2_injector.attn.value_proj.bias", + "backbone.interactions.11.interaction_units_12.branch1to2_injector.attn.output_proj.bias", + "backbone.interactions.11.interaction_units_12.branch1to2_injector.ffn.fc1.bias", + "backbone.interactions.11.interaction_units_12.branch1to2_injector.ffn.dwconv.dwconv.bias", + "backbone.interactions.11.interaction_units_12.branch1to2_injector.ffn.fc2.bias", + "backbone.interactions.11.interaction_units_12.branch1to2_injector.ffn_norm.weight", + "backbone.interactions.11.interaction_units_12.branch1to2_injector.ffn_norm.bias", + "backbone.interactions.11.interaction_units_23.branch2to1_proj.bias", + "backbone.interactions.11.interaction_units_23.branch2to1_injector.ca_gamma", + "backbone.interactions.11.interaction_units_23.branch2to1_injector.cffn_gamma", + "backbone.interactions.11.interaction_units_23.branch2to1_injector.query_norm.weight", + "backbone.interactions.11.interaction_units_23.branch2to1_injector.query_norm.bias", + "backbone.interactions.11.interaction_units_23.branch2to1_injector.feat_norm.weight", + "backbone.interactions.11.interaction_units_23.branch2to1_injector.feat_norm.bias", + "backbone.interactions.11.interaction_units_23.branch2to1_injector.attn.sampling_offsets.bias", + "backbone.interactions.11.interaction_units_23.branch2to1_injector.attn.attention_weights.bias", + "backbone.interactions.11.interaction_units_23.branch2to1_injector.attn.value_proj.bias", + "backbone.interactions.11.interaction_units_23.branch2to1_injector.attn.output_proj.bias", + "backbone.interactions.11.interaction_units_23.branch2to1_injector.ffn.fc1.bias", + "backbone.interactions.11.interaction_units_23.branch2to1_injector.ffn.dwconv.dwconv.bias", + "backbone.interactions.11.interaction_units_23.branch2to1_injector.ffn.fc2.bias", + "backbone.interactions.11.interaction_units_23.branch2to1_injector.ffn_norm.weight", + "backbone.interactions.11.interaction_units_23.branch2to1_injector.ffn_norm.bias", + "backbone.interactions.11.interaction_units_23.branch1to2_proj.bias", + "backbone.interactions.11.interaction_units_23.branch1to2_injector.ca_gamma", + "backbone.interactions.11.interaction_units_23.branch1to2_injector.cffn_gamma", + "backbone.interactions.11.interaction_units_23.branch1to2_injector.query_norm.weight", + "backbone.interactions.11.interaction_units_23.branch1to2_injector.query_norm.bias", + "backbone.interactions.11.interaction_units_23.branch1to2_injector.feat_norm.weight", + "backbone.interactions.11.interaction_units_23.branch1to2_injector.feat_norm.bias", + "backbone.interactions.11.interaction_units_23.branch1to2_injector.attn.sampling_offsets.bias", + "backbone.interactions.11.interaction_units_23.branch1to2_injector.attn.attention_weights.bias", + "backbone.interactions.11.interaction_units_23.branch1to2_injector.attn.value_proj.bias", + "backbone.interactions.11.interaction_units_23.branch1to2_injector.attn.output_proj.bias", + "backbone.interactions.11.interaction_units_23.branch1to2_injector.ffn.fc1.bias", + "backbone.interactions.11.interaction_units_23.branch1to2_injector.ffn.dwconv.dwconv.bias", + "backbone.interactions.11.interaction_units_23.branch1to2_injector.ffn.fc2.bias", + "backbone.interactions.11.interaction_units_23.branch1to2_injector.ffn_norm.weight", + "backbone.interactions.11.interaction_units_23.branch1to2_injector.ffn_norm.bias", + "backbone.merge_branch1.1.weight", + "backbone.merge_branch1.1.bias", + "backbone.merge_branch1.4.weight", + "backbone.merge_branch1.4.bias", + "backbone.merge_branch2.1.weight", + "backbone.merge_branch2.1.bias", + "backbone.merge_branch2.4.weight", + "backbone.merge_branch2.4.bias", + "backbone.merge_branch3.1.weight", + "backbone.merge_branch3.1.bias", + "backbone.merge_branch3.4.weight", + "backbone.merge_branch3.4.bias", + "backbone.fpn1.0.bias", + "backbone.fpn1.1.weight", + "backbone.fpn1.1.bias", + "backbone.fpn1.3.bias", + "backbone.fpn2.0.bias", + "neck.lateral_convs.0.conv.bias", + "neck.lateral_convs.1.conv.bias", + "neck.lateral_convs.2.conv.bias", + "neck.lateral_convs.3.conv.bias", + "neck.fpn_convs.0.conv.bias", + "neck.fpn_convs.1.conv.bias", + "neck.fpn_convs.2.conv.bias", + "neck.fpn_convs.3.conv.bias", + "rpn_head.rpn_conv.bias", + "rpn_head.rpn_cls.bias", + "rpn_head.rpn_reg.bias", + "roi_head.bbox_head.fc_cls.bias", + "roi_head.bbox_head.fc_reg.bias", + "roi_head.bbox_head.shared_fcs.0.bias", + "roi_head.bbox_head.shared_fcs.1.bias", + "roi_head.mask_head.convs.0.conv.bias", + "roi_head.mask_head.convs.1.conv.bias", + "roi_head.mask_head.convs.2.conv.bias", + "roi_head.mask_head.convs.3.conv.bias", + "roi_head.mask_head.upsample.bias", + "roi_head.mask_head.conv_logits.bias" + ], + "lr_scale": 1.0, + "lr": 0.0001, + "weight_decay": 0.0 + } +} +2024-05-31 12:16:36,511 - mmdet - INFO - Automatic scaling of learning rate (LR) has been disabled. +2024-05-31 12:16:37,689 - mmdet - INFO - Start running, work_dir: /mnt/petrelfs/PIIP/mmdetection/work_dirs/mask_rcnn_beit_sbl_1568_1120_672_fpn_1x_coco_bs16 +2024-05-31 12:16:37,689 - mmdet - INFO - Hooks will be executed in the following order: +before_run: +(VERY_HIGH ) StepLrUpdaterHook +(49 ) ToBFloat16HookMMDet +(NORMAL ) DeepspeedCheckpointHook +(LOW ) DeepspeedDistEvalHook +(VERY_LOW ) TextLoggerHook + -------------------- +before_train_epoch: +(VERY_HIGH ) StepLrUpdaterHook +(NORMAL ) DistSamplerSeedHook +(LOW ) IterTimerHook +(LOW ) DeepspeedDistEvalHook +(VERY_LOW ) TextLoggerHook + -------------------- +before_train_iter: +(VERY_HIGH ) StepLrUpdaterHook +(LOW ) IterTimerHook +(LOW ) DeepspeedDistEvalHook + -------------------- +after_train_iter: +(ABOVE_NORMAL) OptimizerHook +(NORMAL ) DeepspeedCheckpointHook +(LOW ) IterTimerHook +(LOW ) DeepspeedDistEvalHook +(VERY_LOW ) TextLoggerHook + -------------------- +after_train_epoch: +(NORMAL ) DeepspeedCheckpointHook +(LOW ) DeepspeedDistEvalHook +(VERY_LOW ) TextLoggerHook + -------------------- +before_val_epoch: +(NORMAL ) DistSamplerSeedHook +(LOW ) IterTimerHook +(VERY_LOW ) TextLoggerHook + -------------------- +before_val_iter: +(LOW ) IterTimerHook + -------------------- +after_val_iter: +(LOW ) IterTimerHook + -------------------- +after_val_epoch: +(VERY_LOW ) TextLoggerHook + -------------------- +after_run: +(VERY_LOW ) TextLoggerHook + -------------------- +2024-05-31 12:16:37,689 - mmdet - INFO - workflow: [('train', 1)], max: 12 epochs +2024-05-31 12:16:37,700 - mmdet - INFO - Checkpoints will be saved to /mnt/petrelfs/PIIP/mmdetection/work_dirs/mask_rcnn_beit_sbl_1568_1120_672_fpn_1x_coco_bs16 by HardDiskBackend. +2024-05-31 12:17:33,278 - mmdet - INFO - Epoch [1][50/7330] lr: 9.890e-06, eta: 1 day, 3:08:17, time: 1.111, data_time: 0.136, memory: 26116, loss_rpn_cls: 0.5883, loss_rpn_bbox: 0.1289, loss_cls: 2.2622, acc: 60.0251, loss_bbox: 0.0239, loss_mask: 1.0976, loss: 4.1009 +2024-05-31 12:18:21,964 - mmdet - INFO - Epoch [1][100/7330] lr: 1.988e-05, eta: 1 day, 1:26:35, time: 0.974, data_time: 0.053, memory: 26220, loss_rpn_cls: 0.3048, loss_rpn_bbox: 0.1046, loss_cls: 0.3776, acc: 95.8550, loss_bbox: 0.1253, loss_mask: 0.7783, loss: 1.6907 +2024-05-31 12:19:10,522 - mmdet - INFO - Epoch [1][150/7330] lr: 2.987e-05, eta: 1 day, 0:50:54, time: 0.971, data_time: 0.056, memory: 26220, loss_rpn_cls: 0.2684, loss_rpn_bbox: 0.1087, loss_cls: 0.3645, acc: 94.9902, loss_bbox: 0.1607, loss_mask: 0.7021, loss: 1.6044 +2024-05-31 12:19:58,363 - mmdet - INFO - Epoch [1][200/7330] lr: 3.986e-05, eta: 1 day, 0:27:24, time: 0.957, data_time: 0.072, memory: 26220, loss_rpn_cls: 0.2525, loss_rpn_bbox: 0.1035, loss_cls: 0.3354, acc: 95.1814, loss_bbox: 0.1545, loss_mask: 0.6820, loss: 1.5278 +2024-05-31 12:20:56,269 - mmdet - INFO - Epoch [1][250/7330] lr: 4.985e-05, eta: 1 day, 1:11:51, time: 1.158, data_time: 0.054, memory: 26220, loss_rpn_cls: 0.2259, loss_rpn_bbox: 0.1010, loss_cls: 0.3218, acc: 95.2703, loss_bbox: 0.1551, loss_mask: 0.6732, loss: 1.4769 +2024-05-31 12:21:43,494 - mmdet - INFO - Epoch [1][300/7330] lr: 5.984e-05, eta: 1 day, 0:49:08, time: 0.944, data_time: 0.043, memory: 26220, loss_rpn_cls: 0.1893, loss_rpn_bbox: 0.0956, loss_cls: 0.3582, acc: 94.5483, loss_bbox: 0.1880, loss_mask: 0.6646, loss: 1.4958 +2024-05-31 12:22:31,474 - mmdet - INFO - Epoch [1][350/7330] lr: 6.983e-05, eta: 1 day, 0:35:50, time: 0.960, data_time: 0.061, memory: 26234, loss_rpn_cls: 0.1686, loss_rpn_bbox: 0.1008, loss_cls: 0.3917, acc: 94.0249, loss_bbox: 0.2058, loss_mask: 0.6530, loss: 1.5200 +2024-05-31 12:23:20,624 - mmdet - INFO - Epoch [1][400/7330] lr: 7.982e-05, eta: 1 day, 0:29:56, time: 0.983, data_time: 0.058, memory: 26311, loss_rpn_cls: 0.1592, loss_rpn_bbox: 0.1031, loss_cls: 0.4515, acc: 92.9375, loss_bbox: 0.2446, loss_mask: 0.6344, loss: 1.5929 +2024-05-31 12:24:10,019 - mmdet - INFO - Epoch [1][450/7330] lr: 8.981e-05, eta: 1 day, 0:25:57, time: 0.988, data_time: 0.058, memory: 26311, loss_rpn_cls: 0.1416, loss_rpn_bbox: 0.1017, loss_cls: 0.4543, acc: 92.7715, loss_bbox: 0.2539, loss_mask: 0.6096, loss: 1.5611 +2024-05-31 12:24:58,526 - mmdet - INFO - Epoch [1][500/7330] lr: 9.980e-05, eta: 1 day, 0:20:01, time: 0.970, data_time: 0.060, memory: 26359, loss_rpn_cls: 0.1282, loss_rpn_bbox: 0.0975, loss_cls: 0.4337, acc: 92.7422, loss_bbox: 0.2527, loss_mask: 0.5825, loss: 1.4946 +2024-05-31 12:25:47,563 - mmdet - INFO - Epoch [1][550/7330] lr: 1.000e-04, eta: 1 day, 0:16:24, time: 0.981, data_time: 0.064, memory: 26428, loss_rpn_cls: 0.1131, loss_rpn_bbox: 0.0948, loss_cls: 0.4478, acc: 92.2668, loss_bbox: 0.2701, loss_mask: 0.5548, loss: 1.4807 +2024-05-31 12:26:35,814 - mmdet - INFO - Epoch [1][600/7330] lr: 1.000e-04, eta: 1 day, 0:11:22, time: 0.965, data_time: 0.043, memory: 26428, loss_rpn_cls: 0.1100, loss_rpn_bbox: 0.0950, loss_cls: 0.4384, acc: 92.2429, loss_bbox: 0.2729, loss_mask: 0.5327, loss: 1.4490 +2024-05-31 12:27:24,103 - mmdet - INFO - Epoch [1][650/7330] lr: 1.000e-04, eta: 1 day, 0:07:03, time: 0.966, data_time: 0.065, memory: 26428, loss_rpn_cls: 0.1081, loss_rpn_bbox: 0.0937, loss_cls: 0.4316, acc: 91.8511, loss_bbox: 0.2843, loss_mask: 0.5229, loss: 1.4405 +2024-05-31 12:28:15,866 - mmdet - INFO - Epoch [1][700/7330] lr: 1.000e-04, eta: 1 day, 0:10:28, time: 1.035, data_time: 0.065, memory: 26428, loss_rpn_cls: 0.1021, loss_rpn_bbox: 0.0960, loss_cls: 0.4359, acc: 91.3984, loss_bbox: 0.3003, loss_mask: 0.5111, loss: 1.4454 +2024-05-31 12:29:04,780 - mmdet - INFO - Epoch [1][750/7330] lr: 1.000e-04, eta: 1 day, 0:07:48, time: 0.978, data_time: 0.055, memory: 26428, loss_rpn_cls: 0.1019, loss_rpn_bbox: 0.0950, loss_cls: 0.4264, acc: 91.7114, loss_bbox: 0.2884, loss_mask: 0.5027, loss: 1.4145 +2024-05-31 12:29:53,930 - mmdet - INFO - Epoch [1][800/7330] lr: 1.000e-04, eta: 1 day, 0:05:46, time: 0.983, data_time: 0.051, memory: 26428, loss_rpn_cls: 0.1007, loss_rpn_bbox: 0.0898, loss_cls: 0.4280, acc: 91.1902, loss_bbox: 0.3084, loss_mask: 0.4888, loss: 1.4157 +2024-05-31 12:30:42,033 - mmdet - INFO - Epoch [1][850/7330] lr: 1.000e-04, eta: 1 day, 0:02:06, time: 0.962, data_time: 0.058, memory: 26440, loss_rpn_cls: 0.0951, loss_rpn_bbox: 0.0896, loss_cls: 0.4118, acc: 91.5347, loss_bbox: 0.2921, loss_mask: 0.4832, loss: 1.3718 +2024-05-31 12:31:31,338 - mmdet - INFO - Epoch [1][900/7330] lr: 1.000e-04, eta: 1 day, 0:00:42, time: 0.986, data_time: 0.062, memory: 26449, loss_rpn_cls: 0.0976, loss_rpn_bbox: 0.0949, loss_cls: 0.4133, acc: 91.0688, loss_bbox: 0.3094, loss_mask: 0.4731, loss: 1.3882 +2024-05-31 12:32:19,781 - mmdet - INFO - Epoch [1][950/7330] lr: 1.000e-04, eta: 23:58:02, time: 0.969, data_time: 0.051, memory: 26449, loss_rpn_cls: 0.0903, loss_rpn_bbox: 0.0888, loss_cls: 0.3961, acc: 91.4194, loss_bbox: 0.2966, loss_mask: 0.4666, loss: 1.3384 +2024-05-31 12:33:08,919 - mmdet - INFO - Exp name: mask_rcnn_beit_sbl_1568_1120_672_fpn_1x_coco_bs16.py +2024-05-31 12:33:08,919 - mmdet - INFO - Epoch [1][1000/7330] lr: 1.000e-04, eta: 23:56:34, time: 0.983, data_time: 0.055, memory: 26449, loss_rpn_cls: 0.0870, loss_rpn_bbox: 0.0886, loss_cls: 0.4084, acc: 90.7405, loss_bbox: 0.3234, loss_mask: 0.4677, loss: 1.3750 +2024-05-31 12:33:58,174 - mmdet - INFO - Epoch [1][1050/7330] lr: 1.000e-04, eta: 23:55:19, time: 0.985, data_time: 0.059, memory: 26449, loss_rpn_cls: 0.0893, loss_rpn_bbox: 0.0884, loss_cls: 0.4184, acc: 90.3738, loss_bbox: 0.3337, loss_mask: 0.4712, loss: 1.4009 +2024-05-31 12:34:47,668 - mmdet - INFO - Epoch [1][1100/7330] lr: 1.000e-04, eta: 23:54:25, time: 0.990, data_time: 0.059, memory: 26449, loss_rpn_cls: 0.0846, loss_rpn_bbox: 0.0895, loss_cls: 0.4159, acc: 90.4163, loss_bbox: 0.3292, loss_mask: 0.4487, loss: 1.3679 +2024-05-31 12:35:46,427 - mmdet - INFO - Epoch [1][1150/7330] lr: 1.000e-04, eta: 1 day, 0:05:11, time: 1.175, data_time: 0.052, memory: 26457, loss_rpn_cls: 0.0853, loss_rpn_bbox: 0.0866, loss_cls: 0.3959, acc: 90.8105, loss_bbox: 0.3175, loss_mask: 0.4508, loss: 1.3361 +2024-05-31 12:36:35,206 - mmdet - INFO - Epoch [1][1200/7330] lr: 1.000e-04, eta: 1 day, 0:02:57, time: 0.976, data_time: 0.049, memory: 26457, loss_rpn_cls: 0.0837, loss_rpn_bbox: 0.0868, loss_cls: 0.4059, acc: 90.1772, loss_bbox: 0.3401, loss_mask: 0.4479, loss: 1.3644 +2024-05-31 12:37:24,093 - mmdet - INFO - Epoch [1][1250/7330] lr: 1.000e-04, eta: 1 day, 0:00:56, time: 0.977, data_time: 0.050, memory: 26457, loss_rpn_cls: 0.0812, loss_rpn_bbox: 0.0875, loss_cls: 0.4093, acc: 90.1699, loss_bbox: 0.3418, loss_mask: 0.4469, loss: 1.3668 +2024-05-31 12:38:12,796 - mmdet - INFO - Epoch [1][1300/7330] lr: 1.000e-04, eta: 23:58:51, time: 0.975, data_time: 0.045, memory: 26457, loss_rpn_cls: 0.0810, loss_rpn_bbox: 0.0802, loss_cls: 0.3883, acc: 90.5415, loss_bbox: 0.3280, loss_mask: 0.4344, loss: 1.3118 +2024-05-31 12:39:02,350 - mmdet - INFO - Epoch [1][1350/7330] lr: 1.000e-04, eta: 23:57:44, time: 0.991, data_time: 0.048, memory: 26457, loss_rpn_cls: 0.0836, loss_rpn_bbox: 0.0909, loss_cls: 0.4013, acc: 89.9229, loss_bbox: 0.3447, loss_mask: 0.4398, loss: 1.3603 +2024-05-31 12:39:51,202 - mmdet - INFO - Epoch [1][1400/7330] lr: 1.000e-04, eta: 23:55:56, time: 0.977, data_time: 0.062, memory: 26457, loss_rpn_cls: 0.0764, loss_rpn_bbox: 0.0862, loss_cls: 0.3746, acc: 90.5095, loss_bbox: 0.3324, loss_mask: 0.4300, loss: 1.2996 +2024-05-31 12:40:39,979 - mmdet - INFO - Epoch [1][1450/7330] lr: 1.000e-04, eta: 23:54:07, time: 0.975, data_time: 0.048, memory: 26457, loss_rpn_cls: 0.0736, loss_rpn_bbox: 0.0866, loss_cls: 0.3684, acc: 90.5605, loss_bbox: 0.3257, loss_mask: 0.4183, loss: 1.2727 +2024-05-31 12:41:28,435 - mmdet - INFO - Epoch [1][1500/7330] lr: 1.000e-04, eta: 23:52:04, time: 0.969, data_time: 0.053, memory: 26457, loss_rpn_cls: 0.0708, loss_rpn_bbox: 0.0838, loss_cls: 0.3739, acc: 90.2881, loss_bbox: 0.3369, loss_mask: 0.4191, loss: 1.2845 +2024-05-31 12:42:16,910 - mmdet - INFO - Epoch [1][1550/7330] lr: 1.000e-04, eta: 23:50:06, time: 0.969, data_time: 0.056, memory: 26457, loss_rpn_cls: 0.0695, loss_rpn_bbox: 0.0789, loss_cls: 0.3663, acc: 90.4072, loss_bbox: 0.3318, loss_mask: 0.4261, loss: 1.2727 +2024-05-31 12:43:08,684 - mmdet - INFO - Epoch [1][1600/7330] lr: 1.000e-04, eta: 23:51:11, time: 1.036, data_time: 0.057, memory: 26457, loss_rpn_cls: 0.0788, loss_rpn_bbox: 0.0810, loss_cls: 0.3755, acc: 90.1667, loss_bbox: 0.3364, loss_mask: 0.4176, loss: 1.2894 +2024-05-31 12:43:57,736 - mmdet - INFO - Epoch [1][1650/7330] lr: 1.000e-04, eta: 23:49:47, time: 0.981, data_time: 0.062, memory: 26458, loss_rpn_cls: 0.0789, loss_rpn_bbox: 0.0901, loss_cls: 0.3889, acc: 89.7749, loss_bbox: 0.3517, loss_mask: 0.4147, loss: 1.3243 +2024-05-31 12:44:46,651 - mmdet - INFO - Epoch [1][1700/7330] lr: 1.000e-04, eta: 23:48:17, time: 0.978, data_time: 0.050, memory: 26458, loss_rpn_cls: 0.0699, loss_rpn_bbox: 0.0851, loss_cls: 0.3673, acc: 90.0942, loss_bbox: 0.3443, loss_mask: 0.4185, loss: 1.2851 +2024-05-31 12:45:36,267 - mmdet - INFO - Epoch [1][1750/7330] lr: 1.000e-04, eta: 23:47:25, time: 0.992, data_time: 0.062, memory: 26458, loss_rpn_cls: 0.0704, loss_rpn_bbox: 0.0830, loss_cls: 0.3691, acc: 89.9868, loss_bbox: 0.3381, loss_mask: 0.4052, loss: 1.2658 +2024-05-31 12:46:25,189 - mmdet - INFO - Epoch [1][1800/7330] lr: 1.000e-04, eta: 23:45:59, time: 0.978, data_time: 0.061, memory: 26458, loss_rpn_cls: 0.0633, loss_rpn_bbox: 0.0813, loss_cls: 0.3773, acc: 89.9922, loss_bbox: 0.3403, loss_mask: 0.4106, loss: 1.2728 +2024-05-31 12:47:14,650 - mmdet - INFO - Epoch [1][1850/7330] lr: 1.000e-04, eta: 23:45:01, time: 0.989, data_time: 0.057, memory: 26458, loss_rpn_cls: 0.0693, loss_rpn_bbox: 0.0815, loss_cls: 0.3641, acc: 90.1606, loss_bbox: 0.3339, loss_mask: 0.4091, loss: 1.2579 +2024-05-31 12:48:03,887 - mmdet - INFO - Epoch [1][1900/7330] lr: 1.000e-04, eta: 23:43:52, time: 0.985, data_time: 0.052, memory: 26458, loss_rpn_cls: 0.0645, loss_rpn_bbox: 0.0831, loss_cls: 0.3706, acc: 89.9077, loss_bbox: 0.3490, loss_mask: 0.4045, loss: 1.2716 +2024-05-31 12:48:53,231 - mmdet - INFO - Epoch [1][1950/7330] lr: 1.000e-04, eta: 23:42:50, time: 0.987, data_time: 0.060, memory: 26458, loss_rpn_cls: 0.0635, loss_rpn_bbox: 0.0771, loss_cls: 0.3556, acc: 90.0466, loss_bbox: 0.3346, loss_mask: 0.3938, loss: 1.2247 +2024-05-31 12:49:42,707 - mmdet - INFO - Exp name: mask_rcnn_beit_sbl_1568_1120_672_fpn_1x_coco_bs16.py +2024-05-31 12:49:42,707 - mmdet - INFO - Epoch [1][2000/7330] lr: 1.000e-04, eta: 23:41:53, time: 0.989, data_time: 0.051, memory: 26458, loss_rpn_cls: 0.0609, loss_rpn_bbox: 0.0812, loss_cls: 0.3519, acc: 90.3396, loss_bbox: 0.3329, loss_mask: 0.3883, loss: 1.2152 +2024-05-31 12:50:43,444 - mmdet - INFO - Epoch [1][2050/7330] lr: 1.000e-04, eta: 23:48:50, time: 1.215, data_time: 0.044, memory: 26458, loss_rpn_cls: 0.0617, loss_rpn_bbox: 0.0764, loss_cls: 0.3511, acc: 90.4482, loss_bbox: 0.3286, loss_mask: 0.3866, loss: 1.2044 +2024-05-31 12:51:32,762 - mmdet - INFO - Epoch [1][2100/7330] lr: 1.000e-04, eta: 23:47:36, time: 0.986, data_time: 0.058, memory: 26458, loss_rpn_cls: 0.0615, loss_rpn_bbox: 0.0769, loss_cls: 0.3516, acc: 90.1467, loss_bbox: 0.3412, loss_mask: 0.3823, loss: 1.2135 +2024-05-31 12:52:21,707 - mmdet - INFO - Epoch [1][2150/7330] lr: 1.000e-04, eta: 23:46:09, time: 0.979, data_time: 0.064, memory: 26458, loss_rpn_cls: 0.0595, loss_rpn_bbox: 0.0788, loss_cls: 0.3519, acc: 90.0234, loss_bbox: 0.3408, loss_mask: 0.3807, loss: 1.2118 +2024-05-31 12:53:11,003 - mmdet - INFO - Epoch [1][2200/7330] lr: 1.000e-04, eta: 23:44:57, time: 0.986, data_time: 0.049, memory: 26458, loss_rpn_cls: 0.0552, loss_rpn_bbox: 0.0785, loss_cls: 0.3358, acc: 90.3088, loss_bbox: 0.3389, loss_mask: 0.3756, loss: 1.1840 +2024-05-31 12:53:59,931 - mmdet - INFO - Epoch [1][2250/7330] lr: 1.000e-04, eta: 23:43:32, time: 0.978, data_time: 0.050, memory: 26458, loss_rpn_cls: 0.0578, loss_rpn_bbox: 0.0796, loss_cls: 0.3512, acc: 89.6785, loss_bbox: 0.3539, loss_mask: 0.3815, loss: 1.2239 +2024-05-31 12:54:48,976 - mmdet - INFO - Epoch [1][2300/7330] lr: 1.000e-04, eta: 23:42:13, time: 0.981, data_time: 0.042, memory: 26458, loss_rpn_cls: 0.0573, loss_rpn_bbox: 0.0826, loss_cls: 0.3312, acc: 90.2749, loss_bbox: 0.3332, loss_mask: 0.3755, loss: 1.1798 +2024-05-31 12:55:38,906 - mmdet - INFO - Epoch [1][2350/7330] lr: 1.000e-04, eta: 23:41:28, time: 0.999, data_time: 0.054, memory: 26458, loss_rpn_cls: 0.0556, loss_rpn_bbox: 0.0773, loss_cls: 0.3398, acc: 89.9260, loss_bbox: 0.3440, loss_mask: 0.3700, loss: 1.1867 +2024-05-31 12:56:27,688 - mmdet - INFO - Epoch [1][2400/7330] lr: 1.000e-04, eta: 23:40:01, time: 0.976, data_time: 0.044, memory: 26458, loss_rpn_cls: 0.0549, loss_rpn_bbox: 0.0779, loss_cls: 0.3357, acc: 89.9841, loss_bbox: 0.3444, loss_mask: 0.3645, loss: 1.1773 +2024-05-31 12:57:17,139 - mmdet - INFO - Epoch [1][2450/7330] lr: 1.000e-04, eta: 23:39:00, time: 0.989, data_time: 0.056, memory: 26458, loss_rpn_cls: 0.0510, loss_rpn_bbox: 0.0755, loss_cls: 0.3370, acc: 89.7981, loss_bbox: 0.3479, loss_mask: 0.3628, loss: 1.1743 +2024-05-31 12:58:05,804 - mmdet - INFO - Epoch [1][2500/7330] lr: 1.000e-04, eta: 23:37:32, time: 0.973, data_time: 0.058, memory: 26458, loss_rpn_cls: 0.0522, loss_rpn_bbox: 0.0764, loss_cls: 0.3199, acc: 90.2756, loss_bbox: 0.3413, loss_mask: 0.3488, loss: 1.1386 +2024-05-31 12:58:57,363 - mmdet - INFO - Epoch [1][2550/7330] lr: 1.000e-04, eta: 23:37:42, time: 1.031, data_time: 0.055, memory: 26458, loss_rpn_cls: 0.0513, loss_rpn_bbox: 0.0767, loss_cls: 0.3180, acc: 90.4863, loss_bbox: 0.3370, loss_mask: 0.3480, loss: 1.1310 +2024-05-31 12:59:46,212 - mmdet - INFO - Epoch [1][2600/7330] lr: 1.000e-04, eta: 23:36:21, time: 0.977, data_time: 0.053, memory: 26458, loss_rpn_cls: 0.0517, loss_rpn_bbox: 0.0715, loss_cls: 0.3215, acc: 90.1570, loss_bbox: 0.3442, loss_mask: 0.3433, loss: 1.1321 +2024-05-31 13:00:35,426 - mmdet - INFO - Epoch [1][2650/7330] lr: 1.000e-04, eta: 23:35:13, time: 0.984, data_time: 0.056, memory: 26458, loss_rpn_cls: 0.0510, loss_rpn_bbox: 0.0790, loss_cls: 0.2992, acc: 90.5857, loss_bbox: 0.3344, loss_mask: 0.3505, loss: 1.1140 +2024-05-31 13:01:25,018 - mmdet - INFO - Epoch [1][2700/7330] lr: 1.000e-04, eta: 23:34:18, time: 0.992, data_time: 0.059, memory: 26458, loss_rpn_cls: 0.0497, loss_rpn_bbox: 0.0777, loss_cls: 0.3136, acc: 90.1208, loss_bbox: 0.3479, loss_mask: 0.3498, loss: 1.1387 +2024-05-31 13:02:14,011 - mmdet - INFO - Epoch [1][2750/7330] lr: 1.000e-04, eta: 23:33:04, time: 0.980, data_time: 0.056, memory: 26458, loss_rpn_cls: 0.0464, loss_rpn_bbox: 0.0741, loss_cls: 0.3034, acc: 90.1438, loss_bbox: 0.3478, loss_mask: 0.3509, loss: 1.1226 +2024-05-31 13:03:03,573 - mmdet - INFO - Epoch [1][2800/7330] lr: 1.000e-04, eta: 23:32:09, time: 0.991, data_time: 0.061, memory: 26458, loss_rpn_cls: 0.0451, loss_rpn_bbox: 0.0736, loss_cls: 0.3243, acc: 89.6328, loss_bbox: 0.3596, loss_mask: 0.3491, loss: 1.1517 +2024-05-31 13:03:52,778 - mmdet - INFO - Epoch [1][2850/7330] lr: 1.000e-04, eta: 23:31:03, time: 0.984, data_time: 0.041, memory: 26458, loss_rpn_cls: 0.0448, loss_rpn_bbox: 0.0701, loss_cls: 0.2969, acc: 90.6221, loss_bbox: 0.3256, loss_mask: 0.3396, loss: 1.0771 +2024-05-31 13:04:42,079 - mmdet - INFO - Epoch [1][2900/7330] lr: 1.000e-04, eta: 23:30:00, time: 0.986, data_time: 0.063, memory: 26458, loss_rpn_cls: 0.0479, loss_rpn_bbox: 0.0743, loss_cls: 0.3012, acc: 90.4543, loss_bbox: 0.3374, loss_mask: 0.3422, loss: 1.1031 +2024-05-31 13:05:42,277 - mmdet - INFO - Epoch [1][2950/7330] lr: 1.000e-04, eta: 23:34:12, time: 1.204, data_time: 0.059, memory: 26458, loss_rpn_cls: 0.0441, loss_rpn_bbox: 0.0744, loss_cls: 0.3094, acc: 90.2610, loss_bbox: 0.3452, loss_mask: 0.3440, loss: 1.1171 +2024-05-31 13:06:31,866 - mmdet - INFO - Exp name: mask_rcnn_beit_sbl_1568_1120_672_fpn_1x_coco_bs16.py +2024-05-31 13:06:31,866 - mmdet - INFO - Epoch [1][3000/7330] lr: 1.000e-04, eta: 23:33:13, time: 0.992, data_time: 0.053, memory: 26458, loss_rpn_cls: 0.0456, loss_rpn_bbox: 0.0736, loss_cls: 0.2983, acc: 90.5178, loss_bbox: 0.3326, loss_mask: 0.3358, loss: 1.0859 +2024-05-31 13:07:21,071 - mmdet - INFO - Epoch [1][3050/7330] lr: 1.000e-04, eta: 23:32:04, time: 0.984, data_time: 0.061, memory: 26458, loss_rpn_cls: 0.0456, loss_rpn_bbox: 0.0711, loss_cls: 0.2939, acc: 90.4861, loss_bbox: 0.3372, loss_mask: 0.3381, loss: 1.0860 +2024-05-31 13:08:09,899 - mmdet - INFO - Epoch [1][3100/7330] lr: 1.000e-04, eta: 23:30:45, time: 0.977, data_time: 0.054, memory: 26458, loss_rpn_cls: 0.0450, loss_rpn_bbox: 0.0700, loss_cls: 0.2919, acc: 90.5027, loss_bbox: 0.3376, loss_mask: 0.3323, loss: 1.0768 +2024-05-31 13:08:59,012 - mmdet - INFO - Epoch [1][3150/7330] lr: 1.000e-04, eta: 23:29:35, time: 0.982, data_time: 0.045, memory: 26458, loss_rpn_cls: 0.0429, loss_rpn_bbox: 0.0744, loss_cls: 0.2908, acc: 90.6816, loss_bbox: 0.3318, loss_mask: 0.3309, loss: 1.0708 +2024-05-31 13:09:48,049 - mmdet - INFO - Epoch [1][3200/7330] lr: 1.000e-04, eta: 23:28:23, time: 0.981, data_time: 0.053, memory: 26458, loss_rpn_cls: 0.0434, loss_rpn_bbox: 0.0689, loss_cls: 0.2887, acc: 90.7112, loss_bbox: 0.3342, loss_mask: 0.3309, loss: 1.0660 +2024-05-31 13:10:36,773 - mmdet - INFO - Epoch [1][3250/7330] lr: 1.000e-04, eta: 23:27:04, time: 0.974, data_time: 0.054, memory: 26458, loss_rpn_cls: 0.0414, loss_rpn_bbox: 0.0685, loss_cls: 0.2935, acc: 90.4656, loss_bbox: 0.3442, loss_mask: 0.3281, loss: 1.0757 +2024-05-31 13:11:26,038 - mmdet - INFO - Epoch [1][3300/7330] lr: 1.000e-04, eta: 23:25:59, time: 0.985, data_time: 0.059, memory: 26458, loss_rpn_cls: 0.0431, loss_rpn_bbox: 0.0685, loss_cls: 0.2912, acc: 90.5679, loss_bbox: 0.3377, loss_mask: 0.3249, loss: 1.0652 +2024-05-31 13:12:15,452 - mmdet - INFO - Epoch [1][3350/7330] lr: 1.000e-04, eta: 23:24:59, time: 0.988, data_time: 0.064, memory: 26458, loss_rpn_cls: 0.0433, loss_rpn_bbox: 0.0697, loss_cls: 0.2970, acc: 90.4829, loss_bbox: 0.3347, loss_mask: 0.3247, loss: 1.0695 +2024-05-31 13:13:04,575 - mmdet - INFO - Epoch [1][3400/7330] lr: 1.000e-04, eta: 23:23:52, time: 0.982, data_time: 0.049, memory: 26458, loss_rpn_cls: 0.0422, loss_rpn_bbox: 0.0689, loss_cls: 0.2778, acc: 90.8223, loss_bbox: 0.3252, loss_mask: 0.3161, loss: 1.0302 +2024-05-31 13:13:56,329 - mmdet - INFO - Epoch [1][3450/7330] lr: 1.000e-04, eta: 23:23:50, time: 1.035, data_time: 0.049, memory: 26458, loss_rpn_cls: 0.0436, loss_rpn_bbox: 0.0672, loss_cls: 0.2853, acc: 90.7161, loss_bbox: 0.3297, loss_mask: 0.3234, loss: 1.0491 +2024-05-31 13:14:45,198 - mmdet - INFO - Epoch [1][3500/7330] lr: 1.000e-04, eta: 23:22:37, time: 0.977, data_time: 0.060, memory: 26458, loss_rpn_cls: 0.0392, loss_rpn_bbox: 0.0666, loss_cls: 0.2802, acc: 90.8262, loss_bbox: 0.3281, loss_mask: 0.3240, loss: 1.0381 +2024-05-31 13:15:34,009 - mmdet - INFO - Epoch [1][3550/7330] lr: 1.000e-04, eta: 23:21:23, time: 0.976, data_time: 0.050, memory: 26458, loss_rpn_cls: 0.0419, loss_rpn_bbox: 0.0676, loss_cls: 0.2755, acc: 90.8887, loss_bbox: 0.3261, loss_mask: 0.3238, loss: 1.0348 +2024-05-31 13:16:23,701 - mmdet - INFO - Epoch [1][3600/7330] lr: 1.000e-04, eta: 23:20:30, time: 0.994, data_time: 0.056, memory: 26458, loss_rpn_cls: 0.0428, loss_rpn_bbox: 0.0711, loss_cls: 0.2890, acc: 90.4236, loss_bbox: 0.3378, loss_mask: 0.3235, loss: 1.0643 +2024-05-31 13:17:12,683 - mmdet - INFO - Epoch [1][3650/7330] lr: 1.000e-04, eta: 23:19:22, time: 0.980, data_time: 0.051, memory: 26458, loss_rpn_cls: 0.0421, loss_rpn_bbox: 0.0696, loss_cls: 0.2869, acc: 90.5654, loss_bbox: 0.3412, loss_mask: 0.3264, loss: 1.0661 +2024-05-31 13:18:01,918 - mmdet - INFO - Epoch [1][3700/7330] lr: 1.000e-04, eta: 23:18:19, time: 0.985, data_time: 0.048, memory: 26458, loss_rpn_cls: 0.0413, loss_rpn_bbox: 0.0670, loss_cls: 0.2765, acc: 90.8862, loss_bbox: 0.3237, loss_mask: 0.3234, loss: 1.0319 +2024-05-31 13:18:51,452 - mmdet - INFO - Epoch [1][3750/7330] lr: 1.000e-04, eta: 23:17:24, time: 0.991, data_time: 0.055, memory: 26458, loss_rpn_cls: 0.0390, loss_rpn_bbox: 0.0665, loss_cls: 0.2714, acc: 91.0356, loss_bbox: 0.3182, loss_mask: 0.3170, loss: 1.0121 +2024-05-31 13:19:40,177 - mmdet - INFO - Epoch [1][3800/7330] lr: 1.000e-04, eta: 23:16:10, time: 0.974, data_time: 0.047, memory: 26458, loss_rpn_cls: 0.0407, loss_rpn_bbox: 0.0630, loss_cls: 0.2723, acc: 91.0955, loss_bbox: 0.3165, loss_mask: 0.3123, loss: 1.0047 +2024-05-31 13:20:33,829 - mmdet - INFO - Epoch [1][3850/7330] lr: 1.000e-04, eta: 23:16:45, time: 1.073, data_time: 0.044, memory: 26458, loss_rpn_cls: 0.0391, loss_rpn_bbox: 0.0676, loss_cls: 0.2763, acc: 90.9138, loss_bbox: 0.3286, loss_mask: 0.3144, loss: 1.0260 +2024-05-31 13:21:28,959 - mmdet - INFO - Epoch [1][3900/7330] lr: 1.000e-04, eta: 23:17:50, time: 1.103, data_time: 0.055, memory: 26458, loss_rpn_cls: 0.0388, loss_rpn_bbox: 0.0674, loss_cls: 0.2790, acc: 90.9868, loss_bbox: 0.3243, loss_mask: 0.3144, loss: 1.0239 +2024-05-31 13:22:18,428 - mmdet - INFO - Epoch [1][3950/7330] lr: 1.000e-04, eta: 23:16:51, time: 0.989, data_time: 0.064, memory: 26458, loss_rpn_cls: 0.0413, loss_rpn_bbox: 0.0687, loss_cls: 0.2859, acc: 90.5361, loss_bbox: 0.3360, loss_mask: 0.3165, loss: 1.0484 +2024-05-31 13:23:07,915 - mmdet - INFO - Exp name: mask_rcnn_beit_sbl_1568_1120_672_fpn_1x_coco_bs16.py +2024-05-31 13:23:07,915 - mmdet - INFO - Epoch [1][4000/7330] lr: 1.000e-04, eta: 23:15:53, time: 0.990, data_time: 0.051, memory: 26458, loss_rpn_cls: 0.0388, loss_rpn_bbox: 0.0685, loss_cls: 0.2827, acc: 90.6885, loss_bbox: 0.3258, loss_mask: 0.3110, loss: 1.0268 +2024-05-31 13:23:57,221 - mmdet - INFO - Epoch [1][4050/7330] lr: 1.000e-04, eta: 23:14:51, time: 0.986, data_time: 0.048, memory: 26458, loss_rpn_cls: 0.0379, loss_rpn_bbox: 0.0660, loss_cls: 0.2903, acc: 90.5288, loss_bbox: 0.3306, loss_mask: 0.3169, loss: 1.0417 +2024-05-31 13:24:46,876 - mmdet - INFO - Epoch [1][4100/7330] lr: 1.000e-04, eta: 23:13:57, time: 0.993, data_time: 0.051, memory: 26458, loss_rpn_cls: 0.0448, loss_rpn_bbox: 0.0708, loss_cls: 0.2870, acc: 90.5437, loss_bbox: 0.3287, loss_mask: 0.3212, loss: 1.0525 +2024-05-31 13:25:35,471 - mmdet - INFO - Epoch [1][4150/7330] lr: 1.000e-04, eta: 23:12:42, time: 0.972, data_time: 0.052, memory: 26458, loss_rpn_cls: 0.0399, loss_rpn_bbox: 0.0636, loss_cls: 0.2713, acc: 91.1631, loss_bbox: 0.3124, loss_mask: 0.3103, loss: 0.9975 +2024-05-31 13:26:24,806 - mmdet - INFO - Epoch [1][4200/7330] lr: 1.000e-04, eta: 23:11:41, time: 0.987, data_time: 0.048, memory: 26458, loss_rpn_cls: 0.0426, loss_rpn_bbox: 0.0643, loss_cls: 0.2813, acc: 90.6782, loss_bbox: 0.3263, loss_mask: 0.3186, loss: 1.0331 +2024-05-31 13:27:14,057 - mmdet - INFO - Epoch [1][4250/7330] lr: 1.000e-04, eta: 23:10:40, time: 0.985, data_time: 0.059, memory: 26458, loss_rpn_cls: 0.0412, loss_rpn_bbox: 0.0667, loss_cls: 0.2828, acc: 90.6289, loss_bbox: 0.3323, loss_mask: 0.3153, loss: 1.0383 +2024-05-31 13:28:03,919 - mmdet - INFO - Epoch [1][4300/7330] lr: 1.000e-04, eta: 23:09:50, time: 0.997, data_time: 0.053, memory: 26458, loss_rpn_cls: 0.0403, loss_rpn_bbox: 0.0657, loss_cls: 0.2710, acc: 90.9695, loss_bbox: 0.3215, loss_mask: 0.3134, loss: 1.0118 +2024-05-31 13:28:55,233 - mmdet - INFO - Epoch [1][4350/7330] lr: 1.000e-04, eta: 23:09:29, time: 1.026, data_time: 0.049, memory: 26458, loss_rpn_cls: 0.0373, loss_rpn_bbox: 0.0649, loss_cls: 0.2726, acc: 91.0754, loss_bbox: 0.3171, loss_mask: 0.3105, loss: 1.0025 +2024-05-31 13:29:44,306 - mmdet - INFO - Epoch [1][4400/7330] lr: 1.000e-04, eta: 23:08:24, time: 0.981, data_time: 0.040, memory: 26458, loss_rpn_cls: 0.0372, loss_rpn_bbox: 0.0611, loss_cls: 0.2698, acc: 91.0618, loss_bbox: 0.3155, loss_mask: 0.3138, loss: 0.9975 +2024-05-31 13:30:34,358 - mmdet - INFO - Epoch [1][4450/7330] lr: 1.000e-04, eta: 23:07:38, time: 1.001, data_time: 0.055, memory: 26458, loss_rpn_cls: 0.0426, loss_rpn_bbox: 0.0689, loss_cls: 0.2833, acc: 90.5439, loss_bbox: 0.3348, loss_mask: 0.3120, loss: 1.0417 +2024-05-31 13:31:23,779 - mmdet - INFO - Epoch [1][4500/7330] lr: 1.000e-04, eta: 23:06:40, time: 0.988, data_time: 0.047, memory: 26458, loss_rpn_cls: 0.0371, loss_rpn_bbox: 0.0642, loss_cls: 0.2606, acc: 91.1836, loss_bbox: 0.3154, loss_mask: 0.3168, loss: 0.9941 +2024-05-31 13:32:13,469 - mmdet - INFO - Epoch [1][4550/7330] lr: 1.000e-04, eta: 23:05:48, time: 0.994, data_time: 0.053, memory: 26458, loss_rpn_cls: 0.0369, loss_rpn_bbox: 0.0640, loss_cls: 0.2697, acc: 91.0396, loss_bbox: 0.3184, loss_mask: 0.3051, loss: 0.9940 +2024-05-31 13:33:03,019 - mmdet - INFO - Epoch [1][4600/7330] lr: 1.000e-04, eta: 23:04:53, time: 0.991, data_time: 0.050, memory: 26458, loss_rpn_cls: 0.0375, loss_rpn_bbox: 0.0682, loss_cls: 0.2691, acc: 91.1406, loss_bbox: 0.3164, loss_mask: 0.3035, loss: 0.9946 +2024-05-31 13:33:52,079 - mmdet - INFO - Epoch [1][4650/7330] lr: 1.000e-04, eta: 23:03:49, time: 0.981, data_time: 0.049, memory: 26458, loss_rpn_cls: 0.0377, loss_rpn_bbox: 0.0641, loss_cls: 0.2705, acc: 91.1833, loss_bbox: 0.3120, loss_mask: 0.3044, loss: 0.9887 +2024-05-31 13:34:41,464 - mmdet - INFO - Epoch [1][4700/7330] lr: 1.000e-04, eta: 23:02:51, time: 0.988, data_time: 0.047, memory: 26458, loss_rpn_cls: 0.0381, loss_rpn_bbox: 0.0630, loss_cls: 0.2569, acc: 91.3328, loss_bbox: 0.3077, loss_mask: 0.3012, loss: 0.9668 +2024-05-31 13:35:30,782 - mmdet - INFO - Epoch [1][4750/7330] lr: 1.000e-04, eta: 23:01:52, time: 0.986, data_time: 0.060, memory: 26458, loss_rpn_cls: 0.0389, loss_rpn_bbox: 0.0623, loss_cls: 0.2623, acc: 91.1763, loss_bbox: 0.3152, loss_mask: 0.3026, loss: 0.9814 +2024-05-31 13:36:30,984 - mmdet - INFO - Epoch [1][4800/7330] lr: 1.000e-04, eta: 23:04:02, time: 1.204, data_time: 0.062, memory: 26458, loss_rpn_cls: 0.0423, loss_rpn_bbox: 0.0626, loss_cls: 0.2724, acc: 91.2693, loss_bbox: 0.3092, loss_mask: 0.3145, loss: 1.0010 +2024-05-31 13:37:19,974 - mmdet - INFO - Epoch [1][4850/7330] lr: 1.000e-04, eta: 23:02:56, time: 0.980, data_time: 0.065, memory: 26458, loss_rpn_cls: 0.0372, loss_rpn_bbox: 0.0634, loss_cls: 0.2594, acc: 91.4478, loss_bbox: 0.3103, loss_mask: 0.3078, loss: 0.9781 +2024-05-31 13:38:09,639 - mmdet - INFO - Epoch [1][4900/7330] lr: 1.000e-04, eta: 23:02:02, time: 0.993, data_time: 0.048, memory: 26458, loss_rpn_cls: 0.0400, loss_rpn_bbox: 0.0650, loss_cls: 0.2693, acc: 90.9670, loss_bbox: 0.3253, loss_mask: 0.3057, loss: 1.0052 +2024-05-31 13:38:58,802 - mmdet - INFO - Epoch [1][4950/7330] lr: 1.000e-04, eta: 23:00:59, time: 0.983, data_time: 0.045, memory: 26458, loss_rpn_cls: 0.0367, loss_rpn_bbox: 0.0630, loss_cls: 0.2597, acc: 91.4092, loss_bbox: 0.3147, loss_mask: 0.3065, loss: 0.9806 +2024-05-31 13:39:48,685 - mmdet - INFO - Exp name: mask_rcnn_beit_sbl_1568_1120_672_fpn_1x_coco_bs16.py +2024-05-31 13:39:48,686 - mmdet - INFO - Epoch [1][5000/7330] lr: 1.000e-04, eta: 23:00:09, time: 0.998, data_time: 0.053, memory: 26458, loss_rpn_cls: 0.0335, loss_rpn_bbox: 0.0611, loss_cls: 0.2536, acc: 91.5046, loss_bbox: 0.3067, loss_mask: 0.2976, loss: 0.9524 +2024-05-31 13:40:37,581 - mmdet - INFO - Epoch [1][5050/7330] lr: 1.000e-04, eta: 22:59:02, time: 0.978, data_time: 0.047, memory: 26458, loss_rpn_cls: 0.0377, loss_rpn_bbox: 0.0637, loss_cls: 0.2682, acc: 91.0796, loss_bbox: 0.3147, loss_mask: 0.3047, loss: 0.9889 +2024-05-31 13:41:27,355 - mmdet - INFO - Epoch [1][5100/7330] lr: 1.000e-04, eta: 22:58:11, time: 0.996, data_time: 0.062, memory: 26458, loss_rpn_cls: 0.0410, loss_rpn_bbox: 0.0648, loss_cls: 0.2611, acc: 91.1479, loss_bbox: 0.3176, loss_mask: 0.3027, loss: 0.9872 +2024-05-31 13:42:16,728 - mmdet - INFO - Epoch [1][5150/7330] lr: 1.000e-04, eta: 22:57:12, time: 0.987, data_time: 0.056, memory: 26458, loss_rpn_cls: 0.0341, loss_rpn_bbox: 0.0626, loss_cls: 0.2632, acc: 91.1116, loss_bbox: 0.3225, loss_mask: 0.3059, loss: 0.9883 +2024-05-31 13:43:06,889 - mmdet - INFO - Epoch [1][5200/7330] lr: 1.000e-04, eta: 22:56:27, time: 1.003, data_time: 0.066, memory: 26458, loss_rpn_cls: 0.0370, loss_rpn_bbox: 0.0649, loss_cls: 0.2609, acc: 91.1843, loss_bbox: 0.3150, loss_mask: 0.3047, loss: 0.9825 +2024-05-31 13:43:58,425 - mmdet - INFO - Epoch [1][5250/7330] lr: 1.000e-04, eta: 22:56:03, time: 1.031, data_time: 0.052, memory: 26458, loss_rpn_cls: 0.0352, loss_rpn_bbox: 0.0584, loss_cls: 0.2466, acc: 91.7659, loss_bbox: 0.3001, loss_mask: 0.3050, loss: 0.9453 +2024-05-31 13:44:47,376 - mmdet - INFO - Epoch [1][5300/7330] lr: 1.000e-04, eta: 22:54:58, time: 0.979, data_time: 0.047, memory: 26458, loss_rpn_cls: 0.0350, loss_rpn_bbox: 0.0614, loss_cls: 0.2609, acc: 91.3127, loss_bbox: 0.3105, loss_mask: 0.3084, loss: 0.9763 +2024-05-31 13:45:36,725 - mmdet - INFO - Epoch [1][5350/7330] lr: 1.000e-04, eta: 22:53:59, time: 0.987, data_time: 0.046, memory: 26458, loss_rpn_cls: 0.0389, loss_rpn_bbox: 0.0608, loss_cls: 0.2604, acc: 91.2285, loss_bbox: 0.3109, loss_mask: 0.3024, loss: 0.9734 +2024-05-31 13:46:26,435 - mmdet - INFO - Epoch [1][5400/7330] lr: 1.000e-04, eta: 22:53:06, time: 0.994, data_time: 0.048, memory: 26458, loss_rpn_cls: 0.0406, loss_rpn_bbox: 0.0631, loss_cls: 0.2596, acc: 91.2722, loss_bbox: 0.3088, loss_mask: 0.3070, loss: 0.9791 +2024-05-31 13:47:15,973 - mmdet - INFO - Epoch [1][5450/7330] lr: 1.000e-04, eta: 22:52:11, time: 0.991, data_time: 0.063, memory: 26458, loss_rpn_cls: 0.0365, loss_rpn_bbox: 0.0645, loss_cls: 0.2604, acc: 91.1531, loss_bbox: 0.3202, loss_mask: 0.3043, loss: 0.9860 +2024-05-31 13:48:05,801 - mmdet - INFO - Epoch [1][5500/7330] lr: 1.000e-04, eta: 22:51:20, time: 0.997, data_time: 0.058, memory: 26458, loss_rpn_cls: 0.0395, loss_rpn_bbox: 0.0647, loss_cls: 0.2695, acc: 90.9995, loss_bbox: 0.3214, loss_mask: 0.3079, loss: 1.0029 +2024-05-31 13:48:55,299 - mmdet - INFO - Epoch [1][5550/7330] lr: 1.000e-04, eta: 22:50:25, time: 0.990, data_time: 0.059, memory: 26458, loss_rpn_cls: 0.0396, loss_rpn_bbox: 0.0656, loss_cls: 0.2682, acc: 91.2058, loss_bbox: 0.3162, loss_mask: 0.3075, loss: 0.9971 +2024-05-31 13:49:44,214 - mmdet - INFO - Epoch [1][5600/7330] lr: 1.000e-04, eta: 22:49:20, time: 0.978, data_time: 0.056, memory: 26458, loss_rpn_cls: 0.0347, loss_rpn_bbox: 0.0637, loss_cls: 0.2460, acc: 91.6135, loss_bbox: 0.3013, loss_mask: 0.3017, loss: 0.9475 +2024-05-31 13:50:34,171 - mmdet - INFO - Epoch [1][5650/7330] lr: 1.000e-04, eta: 22:48:32, time: 0.999, data_time: 0.042, memory: 26458, loss_rpn_cls: 0.0379, loss_rpn_bbox: 0.0632, loss_cls: 0.2637, acc: 91.2852, loss_bbox: 0.3101, loss_mask: 0.3048, loss: 0.9797 +2024-05-31 13:51:34,511 - mmdet - INFO - Epoch [1][5700/7330] lr: 1.000e-04, eta: 22:50:13, time: 1.207, data_time: 0.050, memory: 26458, loss_rpn_cls: 0.0382, loss_rpn_bbox: 0.0648, loss_cls: 0.2639, acc: 91.1323, loss_bbox: 0.3135, loss_mask: 0.3039, loss: 0.9844 +2024-05-31 13:52:24,110 - mmdet - INFO - Epoch [1][5750/7330] lr: 1.000e-04, eta: 22:49:17, time: 0.992, data_time: 0.064, memory: 26458, loss_rpn_cls: 0.0360, loss_rpn_bbox: 0.0617, loss_cls: 0.2604, acc: 91.2073, loss_bbox: 0.3139, loss_mask: 0.3065, loss: 0.9785 +2024-05-31 13:53:12,667 - mmdet - INFO - Epoch [1][5800/7330] lr: 1.000e-04, eta: 22:48:07, time: 0.971, data_time: 0.053, memory: 26458, loss_rpn_cls: 0.0325, loss_rpn_bbox: 0.0542, loss_cls: 0.2404, acc: 91.8831, loss_bbox: 0.2900, loss_mask: 0.2886, loss: 0.9057 +2024-05-31 13:54:01,611 - mmdet - INFO - Epoch [1][5850/7330] lr: 1.000e-04, eta: 22:47:03, time: 0.979, data_time: 0.050, memory: 26458, loss_rpn_cls: 0.0345, loss_rpn_bbox: 0.0570, loss_cls: 0.2512, acc: 91.4868, loss_bbox: 0.3028, loss_mask: 0.2965, loss: 0.9420 +2024-05-31 13:54:50,979 - mmdet - INFO - Epoch [1][5900/7330] lr: 1.000e-04, eta: 22:46:05, time: 0.987, data_time: 0.049, memory: 26458, loss_rpn_cls: 0.0371, loss_rpn_bbox: 0.0646, loss_cls: 0.2644, acc: 91.2156, loss_bbox: 0.3139, loss_mask: 0.2963, loss: 0.9762 +2024-05-31 13:55:40,648 - mmdet - INFO - Epoch [1][5950/7330] lr: 1.000e-04, eta: 22:45:11, time: 0.993, data_time: 0.048, memory: 26458, loss_rpn_cls: 0.0354, loss_rpn_bbox: 0.0634, loss_cls: 0.2555, acc: 91.4541, loss_bbox: 0.3079, loss_mask: 0.2928, loss: 0.9549 +2024-05-31 13:56:29,624 - mmdet - INFO - Exp name: mask_rcnn_beit_sbl_1568_1120_672_fpn_1x_coco_bs16.py +2024-05-31 13:56:29,624 - mmdet - INFO - Epoch [1][6000/7330] lr: 1.000e-04, eta: 22:44:08, time: 0.980, data_time: 0.048, memory: 26458, loss_rpn_cls: 0.0370, loss_rpn_bbox: 0.0582, loss_cls: 0.2664, acc: 91.0750, loss_bbox: 0.3156, loss_mask: 0.3021, loss: 0.9793 +2024-05-31 13:57:18,809 - mmdet - INFO - Epoch [1][6050/7330] lr: 1.000e-04, eta: 22:43:08, time: 0.984, data_time: 0.059, memory: 26458, loss_rpn_cls: 0.0361, loss_rpn_bbox: 0.0623, loss_cls: 0.2650, acc: 91.1785, loss_bbox: 0.3106, loss_mask: 0.2958, loss: 0.9698 +2024-05-31 13:58:08,309 - mmdet - INFO - Epoch [1][6100/7330] lr: 1.000e-04, eta: 22:42:13, time: 0.990, data_time: 0.066, memory: 26458, loss_rpn_cls: 0.0359, loss_rpn_bbox: 0.0632, loss_cls: 0.2607, acc: 91.2229, loss_bbox: 0.3043, loss_mask: 0.2977, loss: 0.9619 +2024-05-31 13:58:57,739 - mmdet - INFO - Epoch [1][6150/7330] lr: 1.000e-04, eta: 22:41:16, time: 0.989, data_time: 0.058, memory: 26458, loss_rpn_cls: 0.0323, loss_rpn_bbox: 0.0611, loss_cls: 0.2686, acc: 91.0493, loss_bbox: 0.3196, loss_mask: 0.2948, loss: 0.9763 +2024-05-31 13:59:49,135 - mmdet - INFO - Epoch [1][6200/7330] lr: 1.000e-04, eta: 22:40:46, time: 1.028, data_time: 0.060, memory: 26458, loss_rpn_cls: 0.0345, loss_rpn_bbox: 0.0590, loss_cls: 0.2522, acc: 91.6013, loss_bbox: 0.2991, loss_mask: 0.2944, loss: 0.9392 +2024-05-31 14:00:38,495 - mmdet - INFO - Epoch [1][6250/7330] lr: 1.000e-04, eta: 22:39:48, time: 0.987, data_time: 0.060, memory: 26458, loss_rpn_cls: 0.0336, loss_rpn_bbox: 0.0590, loss_cls: 0.2465, acc: 91.7048, loss_bbox: 0.3001, loss_mask: 0.2940, loss: 0.9331 +2024-05-31 14:01:28,663 - mmdet - INFO - Epoch [1][6300/7330] lr: 1.000e-04, eta: 22:39:02, time: 1.003, data_time: 0.056, memory: 26458, loss_rpn_cls: 0.0358, loss_rpn_bbox: 0.0626, loss_cls: 0.2647, acc: 91.2305, loss_bbox: 0.3122, loss_mask: 0.3046, loss: 0.9798 +2024-05-31 14:02:18,264 - mmdet - INFO - Epoch [1][6350/7330] lr: 1.000e-04, eta: 22:38:07, time: 0.992, data_time: 0.056, memory: 26458, loss_rpn_cls: 0.0376, loss_rpn_bbox: 0.0624, loss_cls: 0.2566, acc: 91.3335, loss_bbox: 0.3090, loss_mask: 0.2999, loss: 0.9655 +2024-05-31 14:03:07,212 - mmdet - INFO - Epoch [1][6400/7330] lr: 1.000e-04, eta: 22:37:05, time: 0.979, data_time: 0.055, memory: 26458, loss_rpn_cls: 0.0347, loss_rpn_bbox: 0.0596, loss_cls: 0.2534, acc: 91.3765, loss_bbox: 0.3079, loss_mask: 0.2959, loss: 0.9514 +2024-05-31 14:03:56,487 - mmdet - INFO - Epoch [1][6450/7330] lr: 1.000e-04, eta: 22:36:07, time: 0.986, data_time: 0.054, memory: 26458, loss_rpn_cls: 0.0337, loss_rpn_bbox: 0.0619, loss_cls: 0.2492, acc: 91.5754, loss_bbox: 0.2993, loss_mask: 0.2925, loss: 0.9366 +2024-05-31 14:04:46,544 - mmdet - INFO - Epoch [1][6500/7330] lr: 1.000e-04, eta: 22:35:19, time: 1.001, data_time: 0.067, memory: 26458, loss_rpn_cls: 0.0375, loss_rpn_bbox: 0.0667, loss_cls: 0.2607, acc: 91.1309, loss_bbox: 0.3184, loss_mask: 0.3026, loss: 0.9858 +2024-05-31 14:05:35,647 - mmdet - INFO - Epoch [1][6550/7330] lr: 1.000e-04, eta: 22:34:19, time: 0.982, data_time: 0.053, memory: 26458, loss_rpn_cls: 0.0333, loss_rpn_bbox: 0.0576, loss_cls: 0.2392, acc: 91.8489, loss_bbox: 0.2899, loss_mask: 0.2843, loss: 0.9042 +2024-05-31 14:06:35,107 - mmdet - INFO - Epoch [1][6600/7330] lr: 1.000e-04, eta: 22:35:27, time: 1.189, data_time: 0.060, memory: 26458, loss_rpn_cls: 0.0348, loss_rpn_bbox: 0.0586, loss_cls: 0.2448, acc: 91.7166, loss_bbox: 0.2999, loss_mask: 0.2914, loss: 0.9295 +2024-05-31 14:07:24,453 - mmdet - INFO - Epoch [1][6650/7330] lr: 1.000e-04, eta: 22:34:29, time: 0.987, data_time: 0.056, memory: 26458, loss_rpn_cls: 0.0379, loss_rpn_bbox: 0.0624, loss_cls: 0.2500, acc: 91.5134, loss_bbox: 0.3000, loss_mask: 0.2998, loss: 0.9502 +2024-05-31 14:08:13,705 - mmdet - INFO - Epoch [1][6700/7330] lr: 1.000e-04, eta: 22:33:30, time: 0.985, data_time: 0.047, memory: 26458, loss_rpn_cls: 0.0346, loss_rpn_bbox: 0.0586, loss_cls: 0.2396, acc: 91.7891, loss_bbox: 0.2960, loss_mask: 0.2871, loss: 0.9158 +2024-05-31 14:09:02,754 - mmdet - INFO - Epoch [1][6750/7330] lr: 1.000e-04, eta: 22:32:29, time: 0.981, data_time: 0.054, memory: 26458, loss_rpn_cls: 0.0319, loss_rpn_bbox: 0.0606, loss_cls: 0.2434, acc: 91.6177, loss_bbox: 0.3011, loss_mask: 0.2889, loss: 0.9258 +2024-05-31 14:09:52,669 - mmdet - INFO - Epoch [1][6800/7330] lr: 1.000e-04, eta: 22:31:39, time: 0.998, data_time: 0.076, memory: 26458, loss_rpn_cls: 0.0375, loss_rpn_bbox: 0.0636, loss_cls: 0.2571, acc: 91.0654, loss_bbox: 0.3151, loss_mask: 0.2980, loss: 0.9714 +2024-05-31 14:10:42,294 - mmdet - INFO - Epoch [1][6850/7330] lr: 1.000e-04, eta: 22:30:45, time: 0.993, data_time: 0.063, memory: 26458, loss_rpn_cls: 0.0338, loss_rpn_bbox: 0.0622, loss_cls: 0.2450, acc: 91.5815, loss_bbox: 0.3006, loss_mask: 0.2924, loss: 0.9341 +2024-05-31 14:11:31,954 - mmdet - INFO - Epoch [1][6900/7330] lr: 1.000e-04, eta: 22:29:51, time: 0.993, data_time: 0.050, memory: 26458, loss_rpn_cls: 0.0370, loss_rpn_bbox: 0.0614, loss_cls: 0.2670, acc: 91.0227, loss_bbox: 0.3155, loss_mask: 0.2954, loss: 0.9763 +2024-05-31 14:12:20,501 - mmdet - INFO - Epoch [1][6950/7330] lr: 1.000e-04, eta: 22:28:45, time: 0.971, data_time: 0.052, memory: 26458, loss_rpn_cls: 0.0302, loss_rpn_bbox: 0.0528, loss_cls: 0.2409, acc: 91.8813, loss_bbox: 0.2937, loss_mask: 0.2885, loss: 0.9061 +2024-05-31 14:13:10,146 - mmdet - INFO - Exp name: mask_rcnn_beit_sbl_1568_1120_672_fpn_1x_coco_bs16.py +2024-05-31 14:13:10,146 - mmdet - INFO - Epoch [1][7000/7330] lr: 1.000e-04, eta: 22:27:51, time: 0.993, data_time: 0.063, memory: 26458, loss_rpn_cls: 0.0339, loss_rpn_bbox: 0.0626, loss_cls: 0.2532, acc: 91.4211, loss_bbox: 0.3040, loss_mask: 0.2941, loss: 0.9477 +2024-05-31 14:13:59,277 - mmdet - INFO - Epoch [1][7050/7330] lr: 1.000e-04, eta: 22:26:52, time: 0.983, data_time: 0.053, memory: 26458, loss_rpn_cls: 0.0374, loss_rpn_bbox: 0.0628, loss_cls: 0.2505, acc: 91.5173, loss_bbox: 0.3033, loss_mask: 0.2901, loss: 0.9441 +2024-05-31 14:14:50,270 - mmdet - INFO - Epoch [1][7100/7330] lr: 1.000e-04, eta: 22:26:14, time: 1.020, data_time: 0.061, memory: 26458, loss_rpn_cls: 0.0329, loss_rpn_bbox: 0.0598, loss_cls: 0.2477, acc: 91.6748, loss_bbox: 0.2987, loss_mask: 0.2972, loss: 0.9363 +2024-05-31 14:15:39,476 - mmdet - INFO - Epoch [1][7150/7330] lr: 1.000e-04, eta: 22:25:16, time: 0.984, data_time: 0.062, memory: 26458, loss_rpn_cls: 0.0339, loss_rpn_bbox: 0.0584, loss_cls: 0.2466, acc: 91.7141, loss_bbox: 0.2942, loss_mask: 0.2854, loss: 0.9186 +2024-05-31 14:16:29,200 - mmdet - INFO - Epoch [1][7200/7330] lr: 1.000e-04, eta: 22:24:23, time: 0.995, data_time: 0.052, memory: 26458, loss_rpn_cls: 0.0369, loss_rpn_bbox: 0.0614, loss_cls: 0.2541, acc: 91.3611, loss_bbox: 0.3033, loss_mask: 0.2898, loss: 0.9454 +2024-05-31 14:17:18,533 - mmdet - INFO - Epoch [1][7250/7330] lr: 1.000e-04, eta: 22:23:26, time: 0.986, data_time: 0.051, memory: 26458, loss_rpn_cls: 0.0318, loss_rpn_bbox: 0.0582, loss_cls: 0.2594, acc: 91.1836, loss_bbox: 0.3048, loss_mask: 0.2893, loss: 0.9436 +2024-05-31 14:18:08,372 - mmdet - INFO - Epoch [1][7300/7330] lr: 1.000e-04, eta: 22:22:36, time: 0.997, data_time: 0.061, memory: 26458, loss_rpn_cls: 0.0352, loss_rpn_bbox: 0.0600, loss_cls: 0.2551, acc: 91.5564, loss_bbox: 0.2962, loss_mask: 0.2864, loss: 0.9329 +2024-05-31 14:18:38,665 - mmdet - INFO - Saving checkpoint at 1 epochs +2024-05-31 14:20:40,337 - mmdet - INFO - Evaluating bbox... +2024-05-31 14:21:10,223 - mmdet - INFO - + Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.347 + Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=1000 ] = 0.615 + Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=1000 ] = 0.360 + Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=1000 ] = 0.187 + Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=1000 ] = 0.396 + Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=1000 ] = 0.493 + Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.467 + Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=300 ] = 0.467 + Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=1000 ] = 0.467 + Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=1000 ] = 0.265 + Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=1000 ] = 0.524 + Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=1000 ] = 0.634 + +2024-05-31 14:21:10,223 - mmdet - INFO - Evaluating segm... +2024-05-31 14:21:44,117 - mmdet - INFO - + Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.332 + Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=1000 ] = 0.573 + Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=1000 ] = 0.345 + Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=1000 ] = 0.125 + Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=1000 ] = 0.367 + Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=1000 ] = 0.546 + Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.433 + Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=300 ] = 0.433 + Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=1000 ] = 0.433 + Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=1000 ] = 0.210 + Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=1000 ] = 0.491 + Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=1000 ] = 0.633 + +2024-05-31 14:21:44,483 - mmdet - INFO - Exp name: mask_rcnn_beit_sbl_1568_1120_672_fpn_1x_coco_bs16.py +2024-05-31 14:21:44,484 - mmdet - INFO - Epoch(val) [1][625] bbox_mAP: 0.3470, bbox_mAP_50: 0.6150, bbox_mAP_75: 0.3600, bbox_mAP_s: 0.1870, bbox_mAP_m: 0.3960, bbox_mAP_l: 0.4930, bbox_mAP_copypaste: 0.347 0.615 0.360 0.187 0.396 0.493, segm_mAP: 0.3320, segm_mAP_50: 0.5730, segm_mAP_75: 0.3450, segm_mAP_s: 0.1250, segm_mAP_m: 0.3670, segm_mAP_l: 0.5460, segm_mAP_copypaste: 0.332 0.573 0.345 0.125 0.367 0.546 +2024-05-31 14:22:40,629 - mmdet - INFO - Epoch [2][50/7330] lr: 1.000e-04, eta: 22:16:56, time: 1.123, data_time: 0.121, memory: 26458, loss_rpn_cls: 0.0294, loss_rpn_bbox: 0.0564, loss_cls: 0.2345, acc: 91.9177, loss_bbox: 0.2905, loss_mask: 0.2855, loss: 0.8963 +2024-05-31 14:23:29,600 - mmdet - INFO - Epoch [2][100/7330] lr: 1.000e-04, eta: 22:15:58, time: 0.979, data_time: 0.044, memory: 26458, loss_rpn_cls: 0.0302, loss_rpn_bbox: 0.0550, loss_cls: 0.2304, acc: 92.0796, loss_bbox: 0.2835, loss_mask: 0.2791, loss: 0.8782 +2024-05-31 14:24:18,520 - mmdet - INFO - Epoch [2][150/7330] lr: 1.000e-04, eta: 22:14:59, time: 0.978, data_time: 0.044, memory: 26458, loss_rpn_cls: 0.0302, loss_rpn_bbox: 0.0568, loss_cls: 0.2352, acc: 91.9402, loss_bbox: 0.2907, loss_mask: 0.2878, loss: 0.9007 +2024-05-31 14:25:07,654 - mmdet - INFO - Epoch [2][200/7330] lr: 1.000e-04, eta: 22:14:02, time: 0.983, data_time: 0.042, memory: 26458, loss_rpn_cls: 0.0280, loss_rpn_bbox: 0.0559, loss_cls: 0.2302, acc: 91.9336, loss_bbox: 0.2918, loss_mask: 0.2798, loss: 0.8857 +2024-05-31 14:25:57,022 - mmdet - INFO - Epoch [2][250/7330] lr: 1.000e-04, eta: 22:13:08, time: 0.987, data_time: 0.051, memory: 26458, loss_rpn_cls: 0.0321, loss_rpn_bbox: 0.0591, loss_cls: 0.2427, acc: 91.6699, loss_bbox: 0.3031, loss_mask: 0.2890, loss: 0.9260 +2024-05-31 14:26:45,954 - mmdet - INFO - Epoch [2][300/7330] lr: 1.000e-04, eta: 22:12:10, time: 0.979, data_time: 0.047, memory: 26458, loss_rpn_cls: 0.0334, loss_rpn_bbox: 0.0587, loss_cls: 0.2386, acc: 91.8203, loss_bbox: 0.2938, loss_mask: 0.2846, loss: 0.9090 +2024-05-31 14:27:35,721 - mmdet - INFO - Epoch [2][350/7330] lr: 1.000e-04, eta: 22:11:20, time: 0.995, data_time: 0.052, memory: 26458, loss_rpn_cls: 0.0326, loss_rpn_bbox: 0.0590, loss_cls: 0.2367, acc: 91.7634, loss_bbox: 0.2961, loss_mask: 0.2839, loss: 0.9083 +2024-05-31 14:28:24,787 - mmdet - INFO - Epoch [2][400/7330] lr: 1.000e-04, eta: 22:10:24, time: 0.981, data_time: 0.046, memory: 26458, loss_rpn_cls: 0.0290, loss_rpn_bbox: 0.0568, loss_cls: 0.2389, acc: 91.8284, loss_bbox: 0.2914, loss_mask: 0.2759, loss: 0.8920 +2024-05-31 14:29:13,927 - mmdet - INFO - Epoch [2][450/7330] lr: 1.000e-04, eta: 22:09:28, time: 0.983, data_time: 0.059, memory: 26458, loss_rpn_cls: 0.0312, loss_rpn_bbox: 0.0617, loss_cls: 0.2431, acc: 91.3938, loss_bbox: 0.3109, loss_mask: 0.2924, loss: 0.9393 +2024-05-31 14:30:03,243 - mmdet - INFO - Epoch [2][500/7330] lr: 1.000e-04, eta: 22:08:33, time: 0.986, data_time: 0.044, memory: 26458, loss_rpn_cls: 0.0316, loss_rpn_bbox: 0.0548, loss_cls: 0.2390, acc: 91.6440, loss_bbox: 0.3012, loss_mask: 0.2880, loss: 0.9145 +2024-05-31 14:30:52,299 - mmdet - INFO - Epoch [2][550/7330] lr: 1.000e-04, eta: 22:07:37, time: 0.981, data_time: 0.050, memory: 26458, loss_rpn_cls: 0.0309, loss_rpn_bbox: 0.0604, loss_cls: 0.2434, acc: 91.5642, loss_bbox: 0.2974, loss_mask: 0.2924, loss: 0.9245 +2024-05-31 14:31:41,327 - mmdet - INFO - Epoch [2][600/7330] lr: 1.000e-04, eta: 22:06:40, time: 0.980, data_time: 0.044, memory: 26458, loss_rpn_cls: 0.0287, loss_rpn_bbox: 0.0580, loss_cls: 0.2333, acc: 91.8213, loss_bbox: 0.2940, loss_mask: 0.2868, loss: 0.9007 +2024-05-31 14:32:30,168 - mmdet - INFO - Epoch [2][650/7330] lr: 1.000e-04, eta: 22:05:41, time: 0.977, data_time: 0.051, memory: 26458, loss_rpn_cls: 0.0312, loss_rpn_bbox: 0.0588, loss_cls: 0.2448, acc: 91.4851, loss_bbox: 0.3041, loss_mask: 0.2902, loss: 0.9291 +2024-05-31 14:33:18,718 - mmdet - INFO - Epoch [2][700/7330] lr: 1.000e-04, eta: 22:04:40, time: 0.971, data_time: 0.060, memory: 26458, loss_rpn_cls: 0.0305, loss_rpn_bbox: 0.0571, loss_cls: 0.2292, acc: 92.0596, loss_bbox: 0.2858, loss_mask: 0.2831, loss: 0.8858 +2024-05-31 14:34:07,309 - mmdet - INFO - Epoch [2][750/7330] lr: 1.000e-04, eta: 22:03:39, time: 0.972, data_time: 0.051, memory: 26458, loss_rpn_cls: 0.0302, loss_rpn_bbox: 0.0519, loss_cls: 0.2239, acc: 92.2546, loss_bbox: 0.2748, loss_mask: 0.2778, loss: 0.8585 +2024-05-31 14:34:56,405 - mmdet - INFO - Epoch [2][800/7330] lr: 1.000e-04, eta: 22:02:43, time: 0.982, data_time: 0.055, memory: 26458, loss_rpn_cls: 0.0297, loss_rpn_bbox: 0.0555, loss_cls: 0.2320, acc: 91.9209, loss_bbox: 0.2866, loss_mask: 0.2802, loss: 0.8840 +2024-05-31 14:35:45,521 - mmdet - INFO - Epoch [2][850/7330] lr: 1.000e-04, eta: 22:01:48, time: 0.982, data_time: 0.050, memory: 26458, loss_rpn_cls: 0.0335, loss_rpn_bbox: 0.0602, loss_cls: 0.2426, acc: 91.7170, loss_bbox: 0.2971, loss_mask: 0.2839, loss: 0.9172 +2024-05-31 14:36:34,808 - mmdet - INFO - Epoch [2][900/7330] lr: 1.000e-04, eta: 22:00:54, time: 0.986, data_time: 0.053, memory: 26458, loss_rpn_cls: 0.0305, loss_rpn_bbox: 0.0562, loss_cls: 0.2384, acc: 91.7952, loss_bbox: 0.2940, loss_mask: 0.2813, loss: 0.9004 +2024-05-31 14:37:23,544 - mmdet - INFO - Epoch [2][950/7330] lr: 1.000e-04, eta: 21:59:55, time: 0.975, data_time: 0.055, memory: 26458, loss_rpn_cls: 0.0315, loss_rpn_bbox: 0.0589, loss_cls: 0.2477, acc: 91.5183, loss_bbox: 0.2972, loss_mask: 0.2870, loss: 0.9223 +2024-05-31 14:38:23,307 - mmdet - INFO - Epoch [2][1000/7330] lr: 1.000e-04, eta: 22:00:41, time: 1.195, data_time: 0.058, memory: 26458, loss_rpn_cls: 0.0302, loss_rpn_bbox: 0.0542, loss_cls: 0.2466, acc: 91.6304, loss_bbox: 0.2968, loss_mask: 0.2863, loss: 0.9141 +2024-05-31 14:39:19,832 - mmdet - INFO - Epoch [2][1050/7330] lr: 1.000e-04, eta: 22:00:56, time: 1.130, data_time: 0.055, memory: 26458, loss_rpn_cls: 0.0303, loss_rpn_bbox: 0.0586, loss_cls: 0.2530, acc: 91.2498, loss_bbox: 0.3099, loss_mask: 0.2879, loss: 0.9397 +2024-05-31 14:40:08,646 - mmdet - INFO - Epoch [2][1100/7330] lr: 1.000e-04, eta: 21:59:57, time: 0.976, data_time: 0.044, memory: 26458, loss_rpn_cls: 0.0312, loss_rpn_bbox: 0.0555, loss_cls: 0.2465, acc: 91.6699, loss_bbox: 0.2928, loss_mask: 0.3008, loss: 0.9269 +2024-05-31 14:40:57,983 - mmdet - INFO - Epoch [2][1150/7330] lr: 1.000e-04, eta: 21:59:03, time: 0.987, data_time: 0.043, memory: 26458, loss_rpn_cls: 0.0306, loss_rpn_bbox: 0.0545, loss_cls: 0.2288, acc: 92.0447, loss_bbox: 0.2840, loss_mask: 0.2785, loss: 0.8764 +2024-05-31 14:41:47,362 - mmdet - INFO - Epoch [2][1200/7330] lr: 1.000e-04, eta: 21:58:09, time: 0.988, data_time: 0.062, memory: 26458, loss_rpn_cls: 0.0274, loss_rpn_bbox: 0.0547, loss_cls: 0.2286, acc: 92.0815, loss_bbox: 0.2814, loss_mask: 0.2760, loss: 0.8682 +2024-05-31 14:42:36,135 - mmdet - INFO - Epoch [2][1250/7330] lr: 1.000e-04, eta: 21:57:10, time: 0.975, data_time: 0.052, memory: 26458, loss_rpn_cls: 0.0294, loss_rpn_bbox: 0.0566, loss_cls: 0.2317, acc: 91.9822, loss_bbox: 0.2905, loss_mask: 0.2773, loss: 0.8855 +2024-05-31 14:43:25,054 - mmdet - INFO - Epoch [2][1300/7330] lr: 1.000e-04, eta: 21:56:12, time: 0.978, data_time: 0.045, memory: 26458, loss_rpn_cls: 0.0316, loss_rpn_bbox: 0.0563, loss_cls: 0.2419, acc: 91.6936, loss_bbox: 0.2928, loss_mask: 0.2818, loss: 0.9045 +2024-05-31 14:44:14,173 - mmdet - INFO - Epoch [2][1350/7330] lr: 1.000e-04, eta: 21:55:17, time: 0.982, data_time: 0.054, memory: 26458, loss_rpn_cls: 0.0271, loss_rpn_bbox: 0.0561, loss_cls: 0.2332, acc: 91.8752, loss_bbox: 0.2904, loss_mask: 0.2780, loss: 0.8847 +2024-05-31 14:45:05,211 - mmdet - INFO - Epoch [2][1400/7330] lr: 1.000e-04, eta: 21:54:38, time: 1.021, data_time: 0.044, memory: 26458, loss_rpn_cls: 0.0283, loss_rpn_bbox: 0.0543, loss_cls: 0.2328, acc: 92.0474, loss_bbox: 0.2805, loss_mask: 0.2806, loss: 0.8765 +2024-05-31 14:45:54,168 - mmdet - INFO - Epoch [2][1450/7330] lr: 1.000e-04, eta: 21:53:41, time: 0.979, data_time: 0.057, memory: 26458, loss_rpn_cls: 0.0293, loss_rpn_bbox: 0.0558, loss_cls: 0.2298, acc: 92.0840, loss_bbox: 0.2833, loss_mask: 0.2791, loss: 0.8773 +2024-05-31 14:46:42,774 - mmdet - INFO - Epoch [2][1500/7330] lr: 1.000e-04, eta: 21:52:41, time: 0.972, data_time: 0.052, memory: 26458, loss_rpn_cls: 0.0319, loss_rpn_bbox: 0.0552, loss_cls: 0.2274, acc: 92.2527, loss_bbox: 0.2779, loss_mask: 0.2758, loss: 0.8682 +2024-05-31 14:47:32,237 - mmdet - INFO - Epoch [2][1550/7330] lr: 1.000e-04, eta: 21:51:48, time: 0.989, data_time: 0.055, memory: 26458, loss_rpn_cls: 0.0311, loss_rpn_bbox: 0.0581, loss_cls: 0.2472, acc: 91.4363, loss_bbox: 0.3059, loss_mask: 0.2855, loss: 0.9278 +2024-05-31 14:48:21,255 - mmdet - INFO - Epoch [2][1600/7330] lr: 1.000e-04, eta: 21:50:52, time: 0.980, data_time: 0.054, memory: 26458, loss_rpn_cls: 0.0321, loss_rpn_bbox: 0.0558, loss_cls: 0.2206, acc: 92.3840, loss_bbox: 0.2750, loss_mask: 0.2748, loss: 0.8584 +2024-05-31 14:49:09,860 - mmdet - INFO - Epoch [2][1650/7330] lr: 1.000e-04, eta: 21:49:52, time: 0.972, data_time: 0.050, memory: 26458, loss_rpn_cls: 0.0295, loss_rpn_bbox: 0.0555, loss_cls: 0.2329, acc: 92.0098, loss_bbox: 0.2886, loss_mask: 0.2787, loss: 0.8851 +2024-05-31 14:49:58,489 - mmdet - INFO - Epoch [2][1700/7330] lr: 1.000e-04, eta: 21:48:52, time: 0.973, data_time: 0.052, memory: 26458, loss_rpn_cls: 0.0328, loss_rpn_bbox: 0.0568, loss_cls: 0.2334, acc: 92.0093, loss_bbox: 0.2809, loss_mask: 0.2800, loss: 0.8838 +2024-05-31 14:50:48,073 - mmdet - INFO - Epoch [2][1750/7330] lr: 1.000e-04, eta: 21:48:01, time: 0.992, data_time: 0.057, memory: 26458, loss_rpn_cls: 0.0293, loss_rpn_bbox: 0.0579, loss_cls: 0.2364, acc: 91.8096, loss_bbox: 0.2918, loss_mask: 0.2793, loss: 0.8946 +2024-05-31 14:51:36,583 - mmdet - INFO - Epoch [2][1800/7330] lr: 1.000e-04, eta: 21:47:01, time: 0.970, data_time: 0.057, memory: 26458, loss_rpn_cls: 0.0284, loss_rpn_bbox: 0.0540, loss_cls: 0.2358, acc: 91.9521, loss_bbox: 0.2861, loss_mask: 0.2830, loss: 0.8873 +2024-05-31 14:52:25,391 - mmdet - INFO - Epoch [2][1850/7330] lr: 1.000e-04, eta: 21:46:03, time: 0.976, data_time: 0.043, memory: 26458, loss_rpn_cls: 0.0283, loss_rpn_bbox: 0.0557, loss_cls: 0.2278, acc: 92.1782, loss_bbox: 0.2798, loss_mask: 0.2803, loss: 0.8720 +2024-05-31 14:53:14,773 - mmdet - INFO - Epoch [2][1900/7330] lr: 1.000e-04, eta: 21:45:10, time: 0.988, data_time: 0.054, memory: 26458, loss_rpn_cls: 0.0326, loss_rpn_bbox: 0.0587, loss_cls: 0.2401, acc: 91.6008, loss_bbox: 0.3016, loss_mask: 0.2854, loss: 0.9183 +2024-05-31 14:54:04,275 - mmdet - INFO - Epoch [2][1950/7330] lr: 1.000e-04, eta: 21:44:19, time: 0.990, data_time: 0.042, memory: 26458, loss_rpn_cls: 0.0292, loss_rpn_bbox: 0.0559, loss_cls: 0.2314, acc: 92.0547, loss_bbox: 0.2842, loss_mask: 0.2728, loss: 0.8734 +2024-05-31 14:54:53,172 - mmdet - INFO - Epoch [2][2000/7330] lr: 1.000e-04, eta: 21:43:22, time: 0.978, data_time: 0.052, memory: 26458, loss_rpn_cls: 0.0279, loss_rpn_bbox: 0.0557, loss_cls: 0.2178, acc: 92.3677, loss_bbox: 0.2737, loss_mask: 0.2736, loss: 0.8487 +2024-05-31 14:55:48,187 - mmdet - INFO - Epoch [2][2050/7330] lr: 1.000e-04, eta: 21:43:16, time: 1.100, data_time: 0.045, memory: 26458, loss_rpn_cls: 0.0328, loss_rpn_bbox: 0.0575, loss_cls: 0.2310, acc: 92.0806, loss_bbox: 0.2817, loss_mask: 0.2845, loss: 0.8875 +2024-05-31 14:56:48,145 - mmdet - INFO - Epoch [2][2100/7330] lr: 1.000e-04, eta: 21:43:52, time: 1.199, data_time: 0.058, memory: 26458, loss_rpn_cls: 0.0288, loss_rpn_bbox: 0.0501, loss_cls: 0.2218, acc: 92.3958, loss_bbox: 0.2799, loss_mask: 0.2787, loss: 0.8593 +2024-05-31 14:57:37,071 - mmdet - INFO - Epoch [2][2150/7330] lr: 1.000e-04, eta: 21:42:54, time: 0.978, data_time: 0.056, memory: 26458, loss_rpn_cls: 0.0313, loss_rpn_bbox: 0.0584, loss_cls: 0.2375, acc: 91.7192, loss_bbox: 0.2952, loss_mask: 0.2843, loss: 0.9067 +2024-05-31 14:58:25,729 - mmdet - INFO - Epoch [2][2200/7330] lr: 1.000e-04, eta: 21:41:55, time: 0.973, data_time: 0.048, memory: 26458, loss_rpn_cls: 0.0278, loss_rpn_bbox: 0.0542, loss_cls: 0.2281, acc: 92.1050, loss_bbox: 0.2837, loss_mask: 0.2747, loss: 0.8685 +2024-05-31 14:59:15,338 - mmdet - INFO - Epoch [2][2250/7330] lr: 1.000e-04, eta: 21:41:04, time: 0.992, data_time: 0.054, memory: 26458, loss_rpn_cls: 0.0317, loss_rpn_bbox: 0.0580, loss_cls: 0.2369, acc: 91.7759, loss_bbox: 0.2988, loss_mask: 0.2769, loss: 0.9023 +2024-05-31 15:00:04,327 - mmdet - INFO - Epoch [2][2300/7330] lr: 1.000e-04, eta: 21:40:07, time: 0.980, data_time: 0.053, memory: 26458, loss_rpn_cls: 0.0287, loss_rpn_bbox: 0.0568, loss_cls: 0.2243, acc: 92.0696, loss_bbox: 0.2841, loss_mask: 0.2739, loss: 0.8679 +2024-05-31 15:00:52,553 - mmdet - INFO - Epoch [2][2350/7330] lr: 1.000e-04, eta: 21:39:05, time: 0.965, data_time: 0.046, memory: 26458, loss_rpn_cls: 0.0282, loss_rpn_bbox: 0.0563, loss_cls: 0.2231, acc: 92.2639, loss_bbox: 0.2716, loss_mask: 0.2741, loss: 0.8534 +2024-05-31 15:01:41,808 - mmdet - INFO - Epoch [2][2400/7330] lr: 1.000e-04, eta: 21:38:11, time: 0.985, data_time: 0.065, memory: 26458, loss_rpn_cls: 0.0318, loss_rpn_bbox: 0.0587, loss_cls: 0.2323, acc: 92.0278, loss_bbox: 0.2869, loss_mask: 0.2749, loss: 0.8846