add init_cfg with type='pretrained'

configs/barlowtwins/benchmarks/resnet50_8xb32-linear-coslr-100e_in1k.py

@@ -5,7 +5,10 @@ _base_ = [
     '../../_base_/default_runtime.py',
 ]
 
-model = dict(backbone=dict(frozen_stages=4))
+model = dict(
+    backbone=dict(
+        frozen_stages=4,
+        init_cfg=dict(type='Pretrained', checkpoint='', prefix='backbone.')))
 
 # runtime settings
 default_hooks = dict(

configs/beit/benchmarks/beit-base-p16_8xb128-coslr-100e_in1k.py

@@ -23,7 +23,8 @@ model = dict(
         out_type='avg_featmap',
         use_abs_pos_emb=False,
         use_rel_pos_bias=True,
-        use_shared_rel_pos_bias=False),
+        use_shared_rel_pos_bias=False,
+        init_cfg=dict(type='Pretrained', checkpoint='', prefix='backbone.')),
     neck=None,
     head=dict(
         type='LinearClsHead',

configs/beitv2/benchmarks/beit-base-p16_8xb128-coslr-100e_in1k.py

@@ -17,7 +17,8 @@ model = dict(
         out_type='avg_featmap',
         use_abs_pos_emb=False,
         use_rel_pos_bias=True,
-        use_shared_rel_pos_bias=False),
+        use_shared_rel_pos_bias=False,
+        init_cfg=dict(type='Pretrained', checkpoint='', prefix='backbone.')),
     neck=None,
     head=dict(
         type='LinearClsHead',

configs/byol/benchmarks/resnet50_8xb512-linear-coslr-90e_in1k.py

@@ -5,7 +5,10 @@ _base_ = [
     '../../_base_/default_runtime.py',
 ]
 
-model = dict(backbone=dict(frozen_stages=4))
+model = dict(
+    backbone=dict(
+        frozen_stages=4,
+        init_cfg=dict(type='Pretrained', checkpoint='', prefix='backbone.')))
 
 # dataset summary
 train_dataloader = dict(batch_size=512)

configs/cae/benchmarks/beit-base-p16_8xb128-coslr-100e_in1k.py

@@ -74,7 +74,7 @@ model = dict(
         use_abs_pos_emb=True,
         use_rel_pos_bias=True,
         use_shared_rel_pos_bias=False,
-        init_cfg=dict(type='Pretrained', checkpoint='')),
+        init_cfg=dict(type='Pretrained', checkpoint='', prefix='backbone.')),
     neck=None,
     head=dict(
         type='LinearClsHead',

configs/densecl/benchmarks/resnet50_8xb32-linear-steplr-100e_in1k.py

@@ -5,7 +5,10 @@ _base_ = [
     '../../_base_/default_runtime.py',
 ]
 
-model = dict(backbone=dict(frozen_stages=4))
+model = dict(
+    backbone=dict(
+        frozen_stages=4,
+        init_cfg=dict(type='Pretrained', checkpoint='', prefix='backbone.')))
 
 # optimizer
 optim_wrapper = dict(

configs/eva/benchmarks/vit-base-p16_8xb128-coslr-100e_in1k.py

@@ -58,7 +58,7 @@ model = dict(
         drop_path_rate=0.1,
         out_type='avg_featmap',
         final_norm=False,
-        init_cfg=dict(type='Pretrained', checkpoint='')),
+        init_cfg=dict(type='Pretrained', checkpoint='', prefix='backbone.')),
     neck=None,
     head=dict(
         type='LinearClsHead',

configs/eva/benchmarks/vit-base-p16_8xb2048-linear-coslr-100e_in1k.py

@@ -19,7 +19,7 @@ model = dict(
         frozen_stages=12,
         out_type='cls_token',
         final_norm=True,
-        init_cfg=dict(type='Pretrained', checkpoint='')),
+        init_cfg=dict(type='Pretrained', checkpoint='', prefix='backbone.')),
     neck=dict(type='ClsBatchNormNeck', input_features=768),
     head=dict(
         type='VisionTransformerClsHead',

configs/mae/benchmarks/vit-base-p16_8xb128-coslr-100e_in1k.py

@@ -57,7 +57,7 @@ model = dict(
         drop_path_rate=0.1,
         out_type='avg_featmap',
         final_norm=False,
-        init_cfg=dict(type='Pretrained', checkpoint='')),
+        init_cfg=dict(type='Pretrained', checkpoint='', prefix='backbone.')),
     neck=None,
     head=dict(
         type='LinearClsHead',

configs/mae/benchmarks/vit-base-p16_8xb2048-linear-coslr-90e_in1k.py

@@ -20,7 +20,7 @@ model = dict(
         frozen_stages=12,
         out_type='cls_token',
         final_norm=True,
-        init_cfg=dict(type='Pretrained', checkpoint='')),
+        init_cfg=dict(type='Pretrained', checkpoint='', prefix='backbone.')),
     neck=dict(type='ClsBatchNormNeck', input_features=768),
     head=dict(
         type='VisionTransformerClsHead',

configs/mae/benchmarks/vit-huge-p14_32xb8-coslr-50e_in1k-448px.py

@@ -59,7 +59,7 @@ model = dict(
         drop_path_rate=0.3,  # set to 0.3
         out_type='avg_featmap',
         final_norm=False,
-        init_cfg=dict(type='Pretrained', checkpoint='')),
+        init_cfg=dict(type='Pretrained', checkpoint='', prefix='backbone.')),
     neck=None,
     head=dict(
         type='LinearClsHead',

configs/mae/benchmarks/vit-huge-p14_8xb128-coslr-50e_in1k.py

@@ -58,7 +58,7 @@ model = dict(
         drop_path_rate=0.3,  # set to 0.3
         out_type='avg_featmap',
         final_norm=False,
-        init_cfg=dict(type='Pretrained', checkpoint='')),
+        init_cfg=dict(type='Pretrained', checkpoint='', prefix='backbone.')),
     neck=None,
     head=dict(
         type='LinearClsHead',

configs/mae/benchmarks/vit-large-p16_8xb128-coslr-50e_in1k.py

@@ -58,7 +58,7 @@ model = dict(
         drop_path_rate=0.2,  # set to 0.2
         out_type='avg_featmap',
         final_norm=False,
-        init_cfg=dict(type='Pretrained', checkpoint='')),
+        init_cfg=dict(type='Pretrained', checkpoint='', prefix='backbone.')),
     neck=None,
     head=dict(
         type='LinearClsHead',

configs/mae/benchmarks/vit-large-p16_8xb2048-linear-coslr-90e_in1k.py

@@ -20,7 +20,7 @@ model = dict(
         frozen_stages=24,
         out_type='cls_token',
         final_norm=True,
-        init_cfg=dict(type='Pretrained', checkpoint='')),
+        init_cfg=dict(type='Pretrained', checkpoint='', prefix='backbone.')),
     neck=dict(type='ClsBatchNormNeck', input_features=1024),
     head=dict(
         type='VisionTransformerClsHead',

configs/maskfeat/benchmarks/vit-base-p16_8xb256-coslr-100e_in1k.py

@@ -56,7 +56,7 @@ model = dict(
         drop_path_rate=0.1,
         out_type='avg_featmap',
         final_norm=False,
-        init_cfg=dict(type='Pretrained', checkpoint='')),
+        init_cfg=dict(type='Pretrained', checkpoint='', prefix='backbone.')),
     neck=None,
     head=dict(
         type='LinearClsHead',

configs/milan/benchmarks/vit-base-p16_8xb128-coslr-100e_in1k.py

@@ -58,7 +58,7 @@ model = dict(
         drop_path_rate=0.1,
         out_type='avg_featmap',
         final_norm=False,
-        init_cfg=dict(type='Pretrained', checkpoint='')),
+        init_cfg=dict(type='Pretrained', checkpoint='', prefix='backbone.')),
     neck=None,
     head=dict(
         type='LinearClsHead',

configs/milan/benchmarks/vit-base-p16_8xb2048-linear-coslr-100e_in1k.py

@@ -19,7 +19,7 @@ model = dict(
         frozen_stages=12,
         out_type='cls_token',
         final_norm=True,
-        init_cfg=dict(type='Pretrained', checkpoint='')),
+        init_cfg=dict(type='Pretrained', checkpoint='', prefix='backbone.')),
     neck=dict(type='ClsBatchNormNeck', input_features=768),
     head=dict(
         type='VisionTransformerClsHead',

configs/mixmim/benchmarks/mixmim-base_8xb128-coslr-100e_in1k.py

@@ -86,6 +86,10 @@ val_dataloader = dict(
 )
 test_dataloader = val_dataloader
 
+model = dict(
+    backbone=dict(
+        init_cfg=dict(type='Pretrained', checkpoint='', prefix='backbone.')))
+
 # optimizer
 optim_wrapper = dict(
     type='OptimWrapper',

configs/mocov2/benchmarks/resnet50_8xb32-linear-steplr-100e_in1k.py

@@ -5,7 +5,10 @@ _base_ = [
     '../../_base_/default_runtime.py',
 ]
 
-model = dict(backbone=dict(frozen_stages=4))
+model = dict(
+    backbone=dict(
+        frozen_stages=4,
+        init_cfg=dict(type='Pretrained', checkpoint='', prefix='backbone.')))
 
 # optimizer
 optim_wrapper = dict(

configs/mocov3/benchmarks/resnet50_8xb128-linear-coslr-90e_in1k.py

@@ -8,7 +8,11 @@ _base_ = [
 # dataset settings
 train_dataloader = dict(batch_size=128)
 
-model = dict(backbone=dict(frozen_stages=4, norm_eval=True))
+model = dict(
+    backbone=dict(
+        frozen_stages=4,
+        norm_eval=True,
+        init_cfg=dict(type='Pretrained', checkpoint='', prefix='backbone.')))
 
 # optimizer
 optim_wrapper = dict(

configs/mocov3/benchmarks/vit-base-p16_8xb128-linear-coslr-90e_in1k.py

@@ -16,7 +16,8 @@ model = dict(
         patch_size=16,
         stop_grad_conv1=True,
         frozen_stages=12,
-        norm_eval=True),
+        norm_eval=True,
+        init_cfg=dict(type='Pretrained', checkpoint='', prefix='backbone.')),
     head=dict(
         type='VisionTransformerClsHead',
         num_classes=1000,

configs/mocov3/benchmarks/vit-base-p16_8xb64-coslr-150e_in1k.py

@@ -12,7 +12,7 @@ model = dict(
         img_size=224,
         patch_size=16,
         drop_path_rate=0.1,
-    ),
+        init_cfg=dict(type='Pretrained', checkpoint='', prefix='backbone.')),
     neck=None,
     head=dict(
         type='VisionTransformerClsHead',

configs/mocov3/benchmarks/vit-large-p16_8xb64-coslr-100e_in1k.py

@@ -12,7 +12,7 @@ model = dict(
         img_size=224,
         patch_size=16,
         drop_path_rate=0.5,
-    ),
+        init_cfg=dict(type='Pretrained', checkpoint='', prefix='backbone.')),
     neck=None,
     head=dict(
         type='VisionTransformerClsHead',

configs/mocov3/benchmarks/vit-small-p16_8xb128-linear-coslr-90e_in1k.py

@@ -16,7 +16,8 @@ model = dict(
         patch_size=16,
         stop_grad_conv1=True,
         frozen_stages=12,
-        norm_eval=True),
+        norm_eval=True,
+        init_cfg=dict(type='Pretrained', checkpoint='', prefix='backbone.')),
     head=dict(
         type='VisionTransformerClsHead',
         num_classes=1000,

configs/simclr/benchmarks/resnet50_8xb512-linear-coslr-90e_in1k.py

@@ -5,7 +5,10 @@ _base_ = [
     '../../_base_/default_runtime.py',
 ]
 
-model = dict(backbone=dict(frozen_stages=4))
+model = dict(
+    backbone=dict(
+        frozen_stages=4,
+        init_cfg=dict(type='Pretrained', checkpoint='', prefix='backbone.')))
 
 # dataset summary
 train_dataloader = dict(batch_size=512)

configs/simmim/benchmarks/swin-base-w6_8xb256-coslr-100e_in1k-192px.py

@@ -9,7 +9,8 @@ model = dict(
     backbone=dict(
         img_size=192,
         drop_path_rate=0.1,
-        stage_cfgs=dict(block_cfgs=dict(window_size=6))))
+        stage_cfgs=dict(block_cfgs=dict(window_size=6)),
+        init_cfg=dict(type='Pretrained', checkpoint='', prefix='backbone.')))
 
 # optimizer settings
 optim_wrapper = dict(

configs/simmim/benchmarks/swin-base-w7_8xb256-coslr-100e_in1k.py

@@ -52,7 +52,8 @@ model = dict(
     backbone=dict(
         img_size=224,
         drop_path_rate=0.1,
-        stage_cfgs=dict(block_cfgs=dict(window_size=7))))
+        stage_cfgs=dict(block_cfgs=dict(window_size=7)),
+        init_cfg=dict(type='Pretrained', checkpoint='', prefix='backbone.')))
 
 # optimizer settings
 optim_wrapper = dict(

configs/simmim/benchmarks/swin-large-w14_8xb256-coslr-100e_in1k.py

@@ -54,7 +54,8 @@ model = dict(
         img_size=224,
         drop_path_rate=0.2,
         stage_cfgs=dict(block_cfgs=dict(window_size=14)),
-        pad_small_map=True),
+        pad_small_map=True,
+        init_cfg=dict(type='Pretrained', checkpoint='', prefix='backbone.')),
     head=dict(in_channels=1536))
 
 # optimizer settings

configs/simsiam/benchmarks/resnet50_8xb512-linear-coslr-90e_in1k.py

@@ -5,7 +5,10 @@ _base_ = [
     '../../_base_/default_runtime.py',
 ]
 
-model = dict(backbone=dict(frozen_stages=4))
+model = dict(
+    backbone=dict(
+        frozen_stages=4,
+        init_cfg=dict(type='Pretrained', checkpoint='', prefix='backbone.')))
 
 # dataset summary
 train_dataloader = dict(batch_size=512)

configs/spark/benchmarks/convnextv2-tiny_8xb256-coslr-300e_in1k.py

@@ -55,7 +55,7 @@ model = dict(
         drop_path_rate=0.1,
         layer_scale_init_value=0.,
         use_grn=True,
-    ),
+        init_cfg=dict(type='Pretrained', checkpoint='', prefix='backbone.')),
     head=dict(
         type='LinearClsHead',
         num_classes=1000,

configs/spark/benchmarks/resnet50_8xb256-coslr-300e_in1k.py

@@ -48,7 +48,7 @@ model = dict(
     backbone=dict(
         norm_cfg=dict(type='SyncBN', requires_grad=True),
         drop_path_rate=0.05,
-    ),
+        init_cfg=dict(type='Pretrained', checkpoint='', prefix='backbone.')),
     head=dict(
         loss=dict(
             type='LabelSmoothLoss', label_smooth_val=0.1, use_sigmoid=True)),

configs/swav/benchmarks/resnet50_8xb512-linear-coslr-90e_in1k.py

@@ -5,7 +5,10 @@ _base_ = [
     '../../_base_/default_runtime.py',
 ]
 
-model = dict(backbone=dict(frozen_stages=4))
+model = dict(
+    backbone=dict(
+        frozen_stages=4,
+        init_cfg=dict(type='Pretrained', checkpoint='', prefix='backbone.')))
 
 # dataset summary
 train_dataloader = dict(batch_size=512)