only keep one file to set swin transformer v2 model config

mmpretrain/configs/_base_/models/swin_transformer_base.py

@@ -4,7 +4,6 @@ from mmpretrain.models import (CrossEntropyLoss, GlobalAveragePooling,
                                ImageClassifier, LinearClsHead, SwinTransformer)
 
 # model settings
-# Only for evaluation
 model = dict(
     type=ImageClassifier,
     backbone=dict(

mmpretrain/configs/_base_/models/swin_transformer_v2/base_256.py

@@ -1,29 +0,0 @@
-# Copyright (c) OpenMMLab. All rights reserved.
-# This is a BETA new format config file, and the usage may change recently.
-from mmengine.model import ConstantInit, TruncNormalInit
-
-from mmpretrain.models import (CutMix, GlobalAveragePooling, ImageClassifier,
-                               LabelSmoothLoss, LinearClsHead, Mixup,
-                               SwinTransformerV2)
-
-# model settings
-model = dict(
-    type=ImageClassifier,
-    backbone=dict(
-        type=SwinTransformerV2, arch='base', img_size=256, drop_path_rate=0.5),
-    neck=dict(type=GlobalAveragePooling),
-    head=dict(
-        type=LinearClsHead,
-        num_classes=1000,
-        in_channels=1024,
-        init_cfg=None,  # suppress the default init_cfg of LinearClsHead.
-        loss=dict(type=LabelSmoothLoss, label_smooth_val=0.1, mode='original'),
-        cal_acc=False),
-    init_cfg=[
-        dict(type=TruncNormalInit, layer='Linear', std=0.02, bias=0.),
-        dict(type=ConstantInit, layer='LayerNorm', val=1., bias=0.)
-    ],
-    train_cfg=dict(
-        augments=[dict(type=Mixup, alpha=0.8),
-                  dict(type=CutMix, alpha=1.0)]),
-)

mmpretrain/configs/_base_/models/swin_transformer_v2/large_256.py

@@ -1,20 +0,0 @@
-# Copyright (c) OpenMMLab. All rights reserved.
-# This is a BETA new format config file, and the usage may change recently.
-from mmpretrain.models import (CrossEntropyLoss, GlobalAveragePooling,
-                               ImageClassifier, LinearClsHead,
-                               SwinTransformerV2)
-
-# model settings
-# Only for evaluation
-model = dict(
-    type=ImageClassifier,
-    backbone=dict(
-        type=SwinTransformerV2, arch='large', img_size=256,
-        drop_path_rate=0.2),
-    neck=dict(type=GlobalAveragePooling),
-    head=dict(
-        type=LinearClsHead,
-        num_classes=1000,
-        in_channels=1536,
-        loss=dict(type=CrossEntropyLoss, loss_weight=1.0),
-        topk=(1, 5)))

mmpretrain/configs/_base_/models/swin_transformer_v2/large_384.py

@@ -1,20 +0,0 @@
-# Copyright (c) OpenMMLab. All rights reserved.
-# This is a BETA new format config file, and the usage may change recently.
-from mmpretrain.models import (CrossEntropyLoss, GlobalAveragePooling,
-                               ImageClassifier, LinearClsHead,
-                               SwinTransformerV2)
-
-# model settings
-# Only for evaluation
-model = dict(
-    type=ImageClassifier,
-    backbone=dict(
-        type=SwinTransformerV2, arch='large', img_size=384,
-        drop_path_rate=0.2),
-    neck=dict(type=GlobalAveragePooling),
-    head=dict(
-        type=LinearClsHead,
-        num_classes=1000,
-        in_channels=1536,
-        loss=dict(type=CrossEntropyLoss, loss_weight=1.0),
-        topk=(1, 5)))

mmpretrain/configs/_base_/models/swin_transformer_v2/small_256.py

@@ -1,30 +0,0 @@
-# Copyright (c) OpenMMLab. All rights reserved.
-# This is a BETA new format config file, and the usage may change recently.
-from mmengine.model import ConstantInit, TruncNormalInit
-
-from mmpretrain.models import (CutMix, GlobalAveragePooling, ImageClassifier,
-                               LabelSmoothLoss, LinearClsHead, Mixup,
-                               SwinTransformerV2)
-
-# model settings
-model = dict(
-    type=ImageClassifier,
-    backbone=dict(
-        type=SwinTransformerV2, arch='small', img_size=256,
-        drop_path_rate=0.3),
-    neck=dict(type=GlobalAveragePooling),
-    head=dict(
-        type=LinearClsHead,
-        num_classes=1000,
-        in_channels=768,
-        init_cfg=None,  # suppress the default init_cfg of LinearClsHead.
-        loss=dict(type=LabelSmoothLoss, label_smooth_val=0.1, mode='original'),
-        cal_acc=False),
-    init_cfg=[
-        dict(type=TruncNormalInit, layer='Linear', std=0.02, bias=0.),
-        dict(type=ConstantInit, layer='LayerNorm', val=1., bias=0.)
-    ],
-    train_cfg=dict(
-        augments=[dict(type=Mixup, alpha=0.8),
-                  dict(type=CutMix, alpha=1.0)]),
-)

mmpretrain/configs/_base_/models/swin_transformer_v2/tiny_256.py

@@ -1,29 +0,0 @@
-# Copyright (c) OpenMMLab. All rights reserved.
-# This is a BETA new format config file, and the usage may change recently.
-from mmengine.model import ConstantInit, TruncNormalInit
-
-from mmpretrain.models import (CutMix, GlobalAveragePooling, ImageClassifier,
-                               LabelSmoothLoss, LinearClsHead, Mixup,
-                               SwinTransformerV2)
-
-# model settings
-model = dict(
-    type=ImageClassifier,
-    backbone=dict(
-        type=SwinTransformerV2, arch='tiny', img_size=256, drop_path_rate=0.2),
-    neck=dict(type=GlobalAveragePooling),
-    head=dict(
-        type=LinearClsHead,
-        num_classes=1000,
-        in_channels=768,
-        init_cfg=None,  # suppress the default init_cfg of LinearClsHead.
-        loss=dict(type=LabelSmoothLoss, label_smooth_val=0.1, mode='original'),
-        cal_acc=False),
-    init_cfg=[
-        dict(type=TruncNormalInit, layer='Linear', std=0.02, bias=0.),
-        dict(type=ConstantInit, layer='LayerNorm', val=1., bias=0.)
-    ],
-    train_cfg=dict(
-        augments=[dict(type=Mixup, alpha=0.8),
-                  dict(type=CutMix, alpha=1.0)]),
-)

mmpretrain/configs/swin_transformer_v2/swinv2_base_w12_8xb128_in21k_192px.py

@@ -5,7 +5,7 @@ from mmengine.config import read_base
 with read_base():
     from .._base_.datasets.imagenet21k_bs128 import *
     from .._base_.default_runtime import *
-    from .._base_.models.swin_transformer_v2.base_256 import *
+    from .._base_.models.swin_transformer_v2_base import *
     from .._base_.schedules.imagenet_bs1024_adamw_swin import *
 
 # model settings

mmpretrain/configs/swin_transformer_v2/swinv2_base_w16_16xb64_in1k_256px.py

@@ -5,7 +5,7 @@ from mmengine.config import read_base
 with read_base():
     from .._base_.datasets.imagenet_bs64_swin_256 import *
     from .._base_.default_runtime import *
-    from .._base_.models.swin_transformer_v2.base_256 import *
+    from .._base_.models.swin_transformer_v2_base import *
     from .._base_.schedules.imagenet_bs1024_adamw_swin import *
 
 model = dict(backbone=dict(window_size=[16, 16, 16, 8]))

mmpretrain/configs/swin_transformer_v2/swinv2_base_w16_in21k_pre_16xb64_in1k_256px.py

@@ -7,7 +7,7 @@ from mmpretrain.models import ImageClassifier
 with read_base():
     from .._base_.datasets.imagenet_bs64_swin_256 import *
     from .._base_.default_runtime import *
-    from .._base_.models.swin_transformer_v2.base_256 import *
+    from .._base_.models.swin_transformer_v2_base import *
     from .._base_.schedules.imagenet_bs1024_adamw_swin import *
 
 model = dict(

mmpretrain/configs/swin_transformer_v2/swinv2_base_w24_in21k_pre_16xb64_in1k_384px.py

@@ -7,7 +7,7 @@ from mmpretrain.models import ImageClassifier
 with read_base():
     from .._base_.datasets.imagenet_bs64_swin_384 import *
     from .._base_.default_runtime import *
-    from .._base_.models.swin_transformer_v2.base_384 import *
+    from .._base_.models.swin_transformer_v2_base import *
     from .._base_.schedules.imagenet_bs1024_adamw_swin import *
 
 model = dict(

mmpretrain/configs/swin_transformer_v2/swinv2_base_w8_16xb64_in1k_256px.py

@@ -1,9 +1,23 @@
 # Copyright (c) OpenMMLab. All rights reserved.
 # This is a BETA new format config file, and the usage may change recently.
 from mmengine.config import read_base
+from mmengine.model import ConstantInit, TruncNormalInit
+
+from mmpretrain.models import CutMix, Mixup
 
 with read_base():
     from .._base_.datasets.imagenet_bs64_swin_256 import *
     from .._base_.default_runtime import *
-    from .._base_.models.swin_transformer_v2.base_256 import *
+    from .._base_.models.swin_transformer_v2_base import *
     from .._base_.schedules.imagenet_bs1024_adamw_swin import *
+
+# model settings
+model.update(
+    backbone=dict(img_size=256, drop_path_rate=0.5),
+    init_cfg=[
+        dict(type=TruncNormalInit, layer='Linear', std=0.02, bias=0.),
+        dict(type=ConstantInit, layer='LayerNorm', val=1., bias=0.)
+    ],
+    train_cfg=dict(
+        augments=[dict(type=Mixup, alpha=0.8),
+                  dict(type=CutMix, alpha=1.0)]))

mmpretrain/configs/swin_transformer_v2/swinv2_large_w12_8xb128_in21k_192px.py

@@ -5,7 +5,7 @@ from mmengine.config import read_base
 with read_base():
     from .._base_.datasets.imagenet21k_bs128 import *
     from .._base_.default_runtime import *
-    from .._base_.models.swin_transformer_v2.base_256 import *
+    from .._base_.models.swin_transformer_v2_base import *
     from .._base_.schedules.imagenet_bs1024_adamw_swin import *
 
 # model settings

mmpretrain/configs/swin_transformer_v2/swinv2_large_w16_in21k_pre_16xb64_in1k_256px.py

@@ -8,7 +8,7 @@ from mmpretrain.models import ImageClassifier
 with read_base():
     from .._base_.datasets.imagenet_bs64_swin_256 import *
     from .._base_.default_runtime import *
-    from .._base_.models.swin_transformer_v2.large_256 import *
+    from .._base_.models.swin_transformer_v2_base import *
     from .._base_.schedules.imagenet_bs1024_adamw_swin import *
 
 model = dict(

mmpretrain/configs/swin_transformer_v2/swinv2_large_w24_in21k_pre_16xb64_in1k_384px.py

@@ -8,7 +8,7 @@ from mmpretrain.models import ImageClassifier
 with read_base():
     from .._base_.datasets.imagenet_bs64_swin_384 import *
     from .._base_.default_runtime import *
-    from .._base_.models.swin_transformer_v2.large_384 import *
+    from .._base_.models.swin_transformer_v2_base import *
     from .._base_.schedules.imagenet_bs1024_adamw_swin import *
 
 model = dict(

mmpretrain/configs/swin_transformer_v2/swinv2_small_w16_16xb64_in1k_256px.py

@@ -5,7 +5,7 @@ from mmengine.config import read_base
 with read_base():
     from .._base_.datasets.imagenet_bs64_swin_256 import *
     from .._base_.default_runtime import *
-    from .._base_.models.swin_transformer_v2.small_256 import *
+    from .._base_.models.swin_transformer_v2_base import *
     from .._base_.schedules.imagenet_bs1024_adamw_swin import *
 
 model = dict(backbone=dict(window_size=[16, 16, 16, 8]))

mmpretrain/configs/swin_transformer_v2/swinv2_small_w8_16xb64_in1k_256px.py

@@ -1,9 +1,24 @@
 # Copyright (c) OpenMMLab. All rights reserved.
 # This is a BETA new format config file, and the usage may change recently.
 from mmengine.config import read_base
+from mmengine.model import ConstantInit, TruncNormalInit
+
+from mmpretrain.models import CutMix, Mixup
 
 with read_base():
     from .._base_.datasets.imagenet_bs64_swin_256 import *
     from .._base_.default_runtime import *
-    from .._base_.models.swin_transformer_v2.small_256 import *
+    from .._base_.models.swin_transformer_v2_base import *
     from .._base_.schedules.imagenet_bs1024_adamw_swin import *
+
+# model settings
+model.update(
+    backbone=dict(arch='small', img_size=256, drop_path_rate=0.3),
+    head=dict(in_channels=768),
+    init_cfg=[
+        dict(type=TruncNormalInit, layer='Linear', std=0.02, bias=0.),
+        dict(type=ConstantInit, layer='LayerNorm', val=1., bias=0.)
+    ],
+    train_cfg=dict(
+        augments=[dict(type=Mixup, alpha=0.8),
+                  dict(type=CutMix, alpha=1.0)]))

mmpretrain/configs/swin_transformer_v2/swinv2_tiny_w16_16xb64_in1k_256px.py

@@ -5,7 +5,7 @@ from mmengine.config import read_base
 with read_base():
     from .._base_.datasets.imagenet_bs64_swin_256 import *
     from .._base_.default_runtime import *
-    from .._base_.models.swin_transformer_v2.tiny_256 import *
+    from .._base_.models.swin_transformer_v2_base import *
     from .._base_.schedules.imagenet_bs1024_adamw_swin import *
 
 model = dict(backbone=dict(window_size=[16, 16, 16, 8]))

mmpretrain/configs/swin_transformer_v2/swinv2_tiny_w8_16xb64_in1k_256px.py

@@ -1,9 +1,24 @@
 # Copyright (c) OpenMMLab. All rights reserved.
 # This is a BETA new format config file, and the usage may change recently.
 from mmengine.config import read_base
+from mmengine.model import ConstantInit, TruncNormalInit
+
+from mmpretrain.models import CutMix, Mixup
 
 with read_base():
     from .._base_.datasets.imagenet_bs64_swin_256 import *
     from .._base_.default_runtime import *
-    from .._base_.models.swin_transformer_v2.tiny_256 import *
+    from .._base_.models.swin_transformer_v2_base import *
     from .._base_.schedules.imagenet_bs1024_adamw_swin import *
+
+# model settings
+model.update(
+    backbone=dict(arch='tiny', img_size=256, drop_path_rate=0.2),
+    head=dict(in_channels=768),
+    init_cfg=[
+        dict(type=TruncNormalInit, layer='Linear', std=0.02, bias=0.),
+        dict(type=ConstantInit, layer='LayerNorm', val=1., bias=0.)
+    ],
+    train_cfg=dict(
+        augments=[dict(type=Mixup, alpha=0.8),
+                  dict(type=CutMix, alpha=1.0)]))