Expand cls model zoo (#55)

1. expand cls model zoo
2. uniform load_pretrained

configs/classification/cifar10/r50_b128_300e_jpg.py

@@ -2,7 +2,6 @@ _base_ = '../../base.py'
 # model settings
 model = dict(
     type='Classification',
-    pretrained=None,
     backbone=dict(
         type='ResNet',
         depth=50,

configs/classification/cifar10/swintiny_b64_5e_jpg.py

@@ -2,7 +2,6 @@ _base_ = '../../base.py'
 # model settings
 model = dict(
     type='Classification',
-    pretrained=None,
     backbone=dict(
         type='PytorchImageModelWrapper',
         model_name='swin_tiny_patch4_window7_224',

configs/classification/imagenet/hrnet/hrnetw18_b32x8_100e_jpg.py

@@ -7,7 +7,6 @@ log_config = dict(
 # model settings
 model = dict(
     type='Classification',
-    pretrained=None,
     backbone=dict(type='HRNet', arch='w18', multi_scale_output=True),
     neck=dict(type='HRFuseScales', in_channels=(18, 36, 72, 144)),
     head=dict(

configs/classification/imagenet/hrnet/imagenet_hrnetw30_jpg.py

@@ -1,3 +1,5 @@
 _base_ = './hrnetw18_b32x8_100e_jpg.py'
 # model settings
-model = dict(backbone=dict(type='HRNet', arch='w30', multi_scale_output=True))
+model = dict(
+    backbone=dict(type='HRNet', arch='w30', multi_scale_output=True),
+    neck=dict(type='HRFuseScales', in_channels=(30, 60, 120, 240)))

configs/classification/imagenet/hrnet/imagenet_hrnetw32_jpg.py

@@ -1,3 +1,5 @@
 _base_ = './hrnetw18_b32x8_100e_jpg.py'
 # model settings
-model = dict(backbone=dict(type='HRNet', arch='w32', multi_scale_output=True))
+model = dict(
+    backbone=dict(type='HRNet', arch='w32', multi_scale_output=True),
+    neck=dict(type='HRFuseScales', in_channels=(32, 64, 128, 256)))

configs/classification/imagenet/hrnet/imagenet_hrnetw40_jpg.py

@@ -1,3 +1,5 @@
 _base_ = './hrnetw18_b32x8_100e_jpg.py'
 # model settings
-model = dict(backbone=dict(type='HRNet', arch='w40', multi_scale_output=True))
+model = dict(
+    backbone=dict(type='HRNet', arch='w40', multi_scale_output=True),
+    neck=dict(type='HRFuseScales', in_channels=(40, 80, 160, 320)))

configs/classification/imagenet/hrnet/imagenet_hrnetw44_jpg.py

@@ -1,3 +1,5 @@
 _base_ = './hrnetw18_b32x8_100e_jpg.py'
 # model settings
-model = dict(backbone=dict(type='HRNet', arch='w44', multi_scale_output=True))
+model = dict(
+    backbone=dict(type='HRNet', arch='w44', multi_scale_output=True),
+    neck=dict(type='HRFuseScales', in_channels=(44, 88, 176, 352)))

configs/classification/imagenet/hrnet/imagenet_hrnetw48_jpg.py

@@ -1,4 +1,6 @@
 _base_ = './hrnetw18_b32x8_100e_jpg.py'
 
 # model settings
-model = dict(backbone=dict(type='HRNet', arch='w48', multi_scale_output=True))
+model = dict(
+    backbone=dict(type='HRNet', arch='w48', multi_scale_output=True),
+    neck=dict(type='HRFuseScales', in_channels=(48, 96, 192, 384)))

configs/classification/imagenet/hrnet/imagenet_hrnetw64_jpg.py

@@ -1,3 +1,5 @@
 _base_ = './hrnetw18_b32x8_100e_jpg.py'
 # model settings
-model = dict(backbone=dict(type='HRNet', arch='w64', multi_scale_output=True))
+model = dict(
+    backbone=dict(type='HRNet', arch='w64', multi_scale_output=True),
+    neck=dict(type='HRFuseScales', in_channels=(64, 128, 256, 512)))

configs/classification/imagenet/resnet/imagenet_resnet101_jpg.py

@@ -2,18 +2,8 @@ _base_ = './resnet50_b32x8_100e_jpg.py'
 # model settings
 model = dict(
     type='Classification',
-    pretrained=None,
     backbone=dict(
         type='ResNet',
         depth=101,
         out_indices=[4],  # 0: conv-1, x: stage-x
-        norm_cfg=dict(type='BN')),
-    head=dict(
-        type='ClsHead',
-        with_avg_pool=True,
-        in_channels=2048,
-        loss_config=dict(
-            type='CrossEntropyLossWithLabelSmooth',
-            label_smooth=0,
-        ),
-        num_classes=1000))
+        norm_cfg=dict(type='BN')))

configs/classification/imagenet/resnet/imagenet_resnet50_jpg.py

@@ -1,19 +1,8 @@
 _base_ = './resnet50_b32x8_100e_jpg.py'
 # model settings
 model = dict(
-    type='Classification',
-    pretrained=None,
     backbone=dict(
         type='ResNet',
         depth=50,
         out_indices=[4],  # 0: conv-1, x: stage-x
-        norm_cfg=dict(type='BN')),
-    head=dict(
-        type='ClsHead',
-        with_avg_pool=True,
-        in_channels=2048,
-        loss_config=dict(
-            type='CrossEntropyLossWithLabelSmooth',
-            label_smooth=0,
-        ),
-        num_classes=1000))
+        norm_cfg=dict(type='BN')))

configs/classification/imagenet/resnet/resnet50_b32x8_100e_jpg.py

@@ -7,7 +7,6 @@ log_config = dict(
 # model settings
 model = dict(
     type='Classification',
-    pretrained=None,
     backbone=dict(
         type='ResNet',
         depth=50,

configs/classification/imagenet/resnext/resnext50-32x4d_b32x8_100e_jpg.py

@@ -7,7 +7,6 @@ log_config = dict(
 # model settings
 model = dict(
     type='Classification',
-    pretrained=None,
     backbone=dict(
         type='ResNeXt',
         depth=50,

configs/classification/imagenet/swint/swin_tiny_patch4_window7_224_b64x16_300e_jpg.py

@@ -16,7 +16,6 @@ model = dict(
         mode='batch',
         label_smoothing=0.1,
         num_classes=1000),
-    pretrained=None,
     backbone=dict(
         type='PytorchImageModelWrapper',
         model_name='swin_tiny_patch4_window7_224',

configs/classification/imagenet/timm/cait/cait_s24_224.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='cait_s24_224'))

configs/classification/imagenet/timm/cait/cait_xxs24_224.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='cait_xxs24_224'))

configs/classification/imagenet/timm/cait/cait_xxs36_224.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='cait_xxs36_224'))

configs/classification/imagenet/timm/coat/coat_mini.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='coat_mini'))

configs/classification/imagenet/timm/coat/coat_tiny.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='coat_tiny'))

configs/classification/imagenet/timm/convit/convit_base.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='convit_base'))

configs/classification/imagenet/timm/convit/convit_small.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='convit_small'))

configs/classification/imagenet/timm/convit/convit_tiny.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='convit_tiny'))

configs/classification/imagenet/timm/convmixer/convmixer_1024_20_ks9_p14.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='convmixer_1024_20_ks9_p14'))

configs/classification/imagenet/timm/convmixer/convmixer_1536_20.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='convmixer_1536_20'))

configs/classification/imagenet/timm/convmixer/convmixer_768_32.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='convmixer_768_32'))

configs/classification/imagenet/timm/convnext/convnext_base.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='convnext_base'))

configs/classification/imagenet/timm/convnext/convnext_large.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='convnext_large'))

configs/classification/imagenet/timm/convnext/convnext_small.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='convnext_small'))

configs/classification/imagenet/timm/convnext/convnext_tiny.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='convnext_tiny'))

configs/classification/imagenet/timm/crossvit/crossvit_base_240.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='crossvit_base_240'))

configs/classification/imagenet/timm/crossvit/crossvit_small_240.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='crossvit_small_240'))

configs/classification/imagenet/timm/crossvit/crossvit_tiny_240.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='crossvit_tiny_240'))

configs/classification/imagenet/timm/deit/deit_base_distilled_patch16_224.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='deit_base_distilled_patch16_224'))

configs/classification/imagenet/timm/deit/deit_base_patch16_224.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='deit_base_patch16_224'))

configs/classification/imagenet/timm/gmixer/gmixer_24_224.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='gmixer_24_224'))

configs/classification/imagenet/timm/gmlp/gmlp_s16_224.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='gmlp_s16_224'))

configs/classification/imagenet/timm/levit/levit_128.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='levit_128'))

configs/classification/imagenet/timm/levit/levit_192.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='levit_192'))

configs/classification/imagenet/timm/levit/levit_256.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='levit_256'))

configs/classification/imagenet/timm/mlp-mixer/mixer_b16_224.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='mixer_b16_224'))

configs/classification/imagenet/timm/mlp-mixer/mixer_l16_224.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='mixer_l16_224'))

configs/classification/imagenet/timm/mobilevit/mobilevit_s.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='mobilevit_s'))

configs/classification/imagenet/timm/mobilevit/mobilevit_xs.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='mobilevit_xs'))

configs/classification/imagenet/timm/mobilevit/mobilevit_xxs.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='mobilevit_xxs'))

configs/classification/imagenet/timm/nest/jx_nest_base.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='jx_nest_base'))

configs/classification/imagenet/timm/nest/jx_nest_small.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='jx_nest_small'))

configs/classification/imagenet/timm/nest/jx_nest_tiny.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='jx_nest_tiny'))

configs/classification/imagenet/timm/pit/pit_b_distilled_224.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='pit_b_distilled_224'))

configs/classification/imagenet/timm/pit/pit_s_distilled_224.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='pit_s_distilled_224'))

configs/classification/imagenet/timm/poolformer/poolformer_m36.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='poolformer_m36'))

configs/classification/imagenet/timm/poolformer/poolformer_m48.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='poolformer_m48'))

configs/classification/imagenet/timm/poolformer/poolformer_s12.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='poolformer_s12'))

configs/classification/imagenet/timm/poolformer/poolformer_s24.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='poolformer_s24'))

configs/classification/imagenet/timm/poolformer/poolformer_s36.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='poolformer_s36'))

configs/classification/imagenet/timm/resmlp/resmlp_12_distilled_224.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='resmlp_12_distilled_224'))

configs/classification/imagenet/timm/resmlp/resmlp_24_distilled_224.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='resmlp_24_distilled_224'))

configs/classification/imagenet/timm/resmlp/resmlp_36_distilled_224.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='resmlp_36_distilled_224'))

configs/classification/imagenet/timm/resmlp/resmlp_big_24_distilled_224.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='resmlp_big_24_distilled_224'))

configs/classification/imagenet/timm/sequencer/sequencer2d_l.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='sequencer2d_l'))

configs/classification/imagenet/timm/sequencer/sequencer2d_m.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='sequencer2d_m'))

configs/classification/imagenet/timm/sequencer/sequencer2d_s.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='sequencer2d_s'))

configs/classification/imagenet/timm/shuffle_transformer/shuffletrans_base_p4_w7_224.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='shuffletrans_base_p4_w7_224'))

configs/classification/imagenet/timm/shuffle_transformer/shuffletrans_small_p4_w7_224.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='shuffletrans_small_p4_w7_224'))

configs/classification/imagenet/timm/shuffle_transformer/shuffletrans_tiny_p4_w7_224.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='shuffletrans_tiny_p4_w7_224'))

configs/classification/imagenet/timm/swint/dynamic_swin_small_p4_w7_224.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='dynamic_swin_small_p4_w7_224'))

configs/classification/imagenet/timm/swint/dynamic_swin_tiny_p4_w7_224.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='dynamic_swin_tiny_p4_w7_224'))

configs/classification/imagenet/timm/swint/swin_base_patch4_window7_224.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='swin_base_patch4_window7_224'))

configs/classification/imagenet/timm/swint/swin_large_patch4_window7_224.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='swin_large_patch4_window7_224'))

configs/classification/imagenet/timm/timm_config.py

@@ -1,45 +1,48 @@
 _base_ = 'configs/base.py'
+
 log_config = dict(
-    interval=100,
+    interval=10,
     hooks=[dict(type='TextLoggerHook'),
            dict(type='TensorboardLoggerHook')])
 
 # model settings
 model = dict(
     type='Classification',
-    pretrained=None,
+    train_preprocess=['mixUp'],
+    pretrained=True,
+    mixup_cfg=dict(
+        mixup_alpha=0.2,
+        prob=1.0,
+        mode='batch',
+        label_smoothing=0.1,
+        num_classes=1000),
     backbone=dict(
         type='PytorchImageModelWrapper',
-        # model_name='pit_xs_distilled_224',
-        # model_name='swin_small_patch4_window7_224',
-        # model_name='swin_tiny_patch4_window7_224',
-        # model_name='swin_base_patch4_window7_224_in22k',
-        model_name='vit_deit_small_distilled_patch16_224',
-        # model_name = 'vit_deit_small_distilled_patch16_224',
-        # model_name = 'resnet50',
+        model_name='vit_base_patch16_224',
         num_classes=1000,
-        pretrained=True,
     ),
     head=dict(
-        # type='ClsHead', with_avg_pool=True, in_channels=384,
-        # type='ClsHead', with_avg_pool=True, in_channels=768,
-        # type='ClsHead', with_avg_pool=True, in_channels=1024,
-        # num_classes=0)
         type='ClsHead',
+        loss_config={
+            'type': 'SoftTargetCrossEntropy',
+        },
         with_fc=False))
 
-data_train_list = 'data/imagenet_raw/meta/train_labeled.txt'
-data_train_root = 'data/imagenet_raw/train/'
-data_test_list = 'data/imagenet_raw/meta/val_labeled.txt'
-data_test_root = 'data/imagenet_raw/validation/'
-data_all_list = 'data/imagenet_raw/meta/all_labeled.txt'
-data_root = 'data/imagenet_raw/'
+load_from = None
+
+data_train_list = './imagenet_raw/meta/train_labeled.txt'
+data_train_root = './imagenet_raw/train/'
+data_test_list = './imagenet_raw/meta/val_labeled.txt'
+data_test_root = './imagenet_raw/val/'
+data_all_list = './imagenet_raw/meta/all_labeled.txt'
+data_root = './imagenet_raw/'
 
 dataset_type = 'ClsDataset'
 img_norm_cfg = dict(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
 train_pipeline = [
     dict(type='RandomResizedCrop', size=224),
     dict(type='RandomHorizontalFlip'),
+    dict(type='MMAutoAugment'),
     dict(type='ToTensor'),
     dict(type='Normalize', **img_norm_cfg),
     dict(type='Collect', keys=['img', 'gt_labels'])
@@ -53,7 +56,7 @@ test_pipeline = [
 ]
 
 data = dict(
-    imgs_per_gpu=32,  # total 256
+    imgs_per_gpu=64,  # total 256
     workers_per_gpu=8,
     train=dict(
         type=dataset_type,
@@ -84,11 +87,25 @@ eval_pipelines = [
 custom_hooks = []
 
 # optimizer
-optimizer = dict(type='SGD', lr=0.1, momentum=0.9, weight_decay=0.0001)
+optimizer = dict(
+    type='AdamW',
+    lr=0.003,
+    weight_decay=0.3,
+    paramwise_options={
+        'cls_token': dict(weight_decay=0.),
+        'pos_embed': dict(weight_decay=0.),
+    })
+optimizer_config = dict(grad_clip=dict(max_norm=1.0), update_interval=8)
 
 # learning policy
-lr_config = dict(policy='step', step=[30, 60, 90])
-checkpoint_config = dict(interval=10)
+lr_config = dict(
+    policy='CosineAnnealing',
+    min_lr=0,
+    warmup='linear',
+    warmup_iters=10000,
+    warmup_ratio=1e-4,
+)
+checkpoint_config = dict(interval=30)
 
 # runtime settings
 total_epochs = 90

configs/classification/imagenet/timm/tnt/tnt_s_patch16_224.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='tnt_s_patch16_224'))

configs/classification/imagenet/timm/twins/twins_svt_base.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='twins_svt_base'))

configs/classification/imagenet/timm/twins/twins_svt_large.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='twins_svt_large'))

configs/classification/imagenet/timm/twins/twins_svt_small.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='twins_svt_small'))

configs/classification/imagenet/timm/vit/vit_base_patch16_224.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='vit_base_patch16_224'))

configs/classification/imagenet/timm/vit/vit_large_patch16_224.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='vit_large_patch16_224'))

configs/classification/imagenet/timm/xcit/xcit_large_24_p8_224.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='xcit_large_24_p8_224'))

configs/classification/imagenet/timm/xcit/xcit_large_24_p8_224_dist.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='xcit_large_24_p8_224_dist'))

configs/classification/imagenet/timm/xcit/xcit_medium_24_p8_224.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='xcit_medium_24_p8_224'))

configs/classification/imagenet/timm/xcit/xcit_medium_24_p8_224_dist.py

@@ -0,0 +1,4 @@
+_base_ = '../timm_config.py'
+
+# model settings
+model = dict(backbone=dict(model_name='xcit_medium_24_p8_224_dist'))

configs/classification/imagenet/vit/vit_base_patch16_224_b64x64_300e_jpg.py

@@ -15,7 +15,6 @@ model = dict(
         mode='batch',
         label_smoothing=0.1,
         num_classes=1000),
-    pretrained=None,
     backbone=dict(
         type='PytorchImageModelWrapper',
         model_name='vit_base_patch16_224',

configs/config_templates/classification.py

@@ -12,7 +12,6 @@ export = dict(export_neck=True)
 # model settings
 model = dict(
     type='Classification',
-    pretrained=None,
     backbone=dict(
         type='ResNet',
         depth=50,

configs/config_templates/classification_oss.py

@@ -19,7 +19,6 @@ export = dict(export_neck=True)
 # model settings
 model = dict(
     type='Classification',
-    pretrained=None,
     backbone=dict(
         type='PytorchImageModelWrapper',
         # model_name='pit_xs_distilled_224',

configs/config_templates/classification_tfrecord_oss.py

@@ -19,7 +19,6 @@ export = dict(export_neck=True)
 # model settings
 model = dict(
     type='Classification',
-    pretrained=None,
     backbone=dict(
         type='ResNet',
         depth=50,

configs/config_templates/metric_learning/modelparallel_softmaxbased_tfrecord_oss.py

@@ -44,7 +44,6 @@ work_dir = 'oss://path/to/work_dirs/classification/'
 # model settings
 model = dict(
     type='Classification',
-    # pretrained=None,
     backbone=dict(
         type='PytorchImageModelWrapper',
 

configs/config_templates/metric_learning/softmaxbased_tfrecord_oss.py

@@ -49,7 +49,6 @@ work_dir = 'oss://path/to/work_dirs/classification/'
 # model settings
 model = dict(
     type='Classification',
-    # pretrained=None,
     backbone=dict(
         type='PytorchImageModelWrapper',
 

configs/metric_learning/cub_resnet50_jpg.py

@@ -7,7 +7,6 @@ log_config = dict(
 # model settings
 model = dict(
     type='Classification',
-    pretrained=None,
     backbone=dict(
         type='ResNet',
         depth=50,

configs/metric_learning/imagenet_resnet50_1000kid_jpg.py

@@ -7,7 +7,6 @@ log_config = dict(
 # model settings
 model = dict(
     type='Classification',
-    pretrained=None,
     backbone=dict(
         type='ResNet',
         depth=50,

configs/metric_learning/sop_timm_swinb_local.py

@@ -13,9 +13,9 @@ load_from = None
 # model settings
 model = dict(
     type='Classification',
+    train_preprocess=['randomErasing'],
     pretrained=
     'http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/pretrained_models/timm/swin_base_patch4_window7_224_22k_statedict.pth',
-    train_preprocess=['randomErasing'],
     backbone=dict(
         type='PytorchImageModelWrapper',
         model_name='swin_base_patch4_window7_224_in22k'

docs/source/model_zoo_cls.md

@@ -2,22 +2,77 @@
 
 ## Benchmarks
 
-| Algorithm | Config                                           | Top-1 (%) | Top-5 (%) | Download                                                     |
-| --------- | ------------------------------------------------------------ | --------- | --------- | ------------------------------------------------------------ |
-| resnet50(raw) | [resnet50(raw)](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/resnet/imagenet_resnet50_jpg.py) | 76.454    | 93.084    | [model](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/resnet/resnet50/epoch_100.pth) |
-| resnet50(tfrecord) | [resnet50(tfrecord)](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/resnet/imagenet_rn50_tfrecord.py) | 76.266    | 92.972    | [model](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/resnet/resnet50/epoch_100.pth) |
-| resnet101 | [resnet101](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/resnet/imagenet_resnet101_jpg.py) | 78.152    | 93.922    | [model](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/resnet/resnet101/epoch_100.pth) |
-| resnet152 | [resnet152](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/resnet/imagenet_resnet152_jpg.py) | 78.544    | 94.206    | [model](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/resnet/resnet152/epoch_100.pth) |
-| resnext50-32x4d | [resnext50-32x4d](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/resnext/imagenet_resnext50-32x4d_jpg.py) | 77.604    | 93.856    | [model](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/resnext/resnet50/epoch_100.pth) |
-| resnext101-32x4d | [resnext101-32x4d](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/resnext/imagenet_resnext101-32x4d_jpg.py) | 78.568    | 94.344    | [model](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/resnext/resnext50-32x4d/epoch_100.pth) |
-| resnext101-32x8d | [resnext101-32x8d](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/resnext/imagenet_resnext101-32x8d_jpg.py) | 79.468    | 94.434    | [model](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/resnext/resnext101-32x8d/epoch_100.pth) |
-| resnext152-32x4d | [resnext152-32x4d](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/resnext/imagenet_resnext152-32x4d_jpg.py) | 78.994    | 94.462    | [model](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/resnext/resnext152-32x4d/epoch_100.pth) |
-| hrnetw18 | [hrnetw18](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/hrnet/imagenet_hrnetw18_jpg.py) | 76.258    | 92.976    | [model](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/hrnet/hrnetw18/epoch_100.pth) |
-| hrnetw30 | [hrnetw30](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/hrnet/imagenet_hrnetw30_jpg.py) | 77.66    | 93.862    | [model](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/hrnet/hrnetw30/epoch_100.pth) |
-| hrnetw32 | [hrnetw32](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/hrnet/imagenet_hrnetw32_jpg.py) | 77.994    | 93.976    | [model](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/hrnet/hrnetw32/epoch_100.pth) |
-| hrnetw40 | [hrnetw40](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/hrnet/imagenet_hrnetw40_jpg.py) | 78.142    | 93.956    | [model](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/hrnet/hrnetw40/epoch_100.pth) |
-| hrnetw44 | [hrnetw44](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/hrnet/imagenet_hrnetw44_jpg.py) | 79.266    | 94.476    | [model](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/hrnet/hrnetw44/epoch_100.pth) |
-| hrnetw48 | [hrnetw48](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/hrnet/imagenet_hrnetw48_jpg.py) | 79.636    | 94.802    | [model](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/hrnet/hrnetw48/epoch_100.pth) |
-| hrnetw64 | [hrnetw64](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/hrnet/imagenet_hrnetw64_jpg.py) | 79.884    | 95.04    | [model](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/resnet/hrnetw64/epoch_100.pth) |
-| vit-base-patch16 | [vit-base-patch16](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/vit/imagenet_vit_base_patch16_224_jpg.py) | 76.082    | 92.026    | [model](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/vit/vit-base-patch16/epoch_300.pth) |
-| swin-tiny-patch4-window7 | [swin-tiny-patch4-window7](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/swint/imagenet_swin_tiny_patch4_window7_224_jpg.py) | 80.528    | 94.822    | [model](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/swint/swin-tiny-patch4-window7/epoch_300.pth) |
+| Algorithm | Config                                           | Top-1 (%) | Top-5 (%) | gpu memory (MB) | inference time (ms/img) | Download                                                     |
+| --------- | ------------------------------------------------------------ | --------- | --------- | --------- | --------- | ------------------------------------------------------------ |
+| resnet50(raw) | [resnet50(raw)](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/resnet/imagenet_resnet50_jpg.py) | 76.454    | 93.084    | 2412    | 8.59    | [model](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/resnet/resnet50/epoch_100.pth) |
+| resnet50(tfrecord) | [resnet50(tfrecord)](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/resnet/imagenet_rn50_tfrecord.py) | 76.266    | 92.972    | 2412    | 8.59    | [model](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/resnet/resnet50/epoch_100.pth) |
+| resnet101 | [resnet101](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/resnet/imagenet_resnet101_jpg.py) | 78.152    | 93.922    | 2484    | 16.77    | [model](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/resnet/resnet101/epoch_100.pth) |
+| resnet152 | [resnet152](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/resnet/imagenet_resnet152_jpg.py) | 78.544    | 94.206    | 2544    | 24.69    | [model](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/resnet/resnet152/epoch_100.pth) |
+| resnext50-32x4d | [resnext50-32x4d](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/resnext/imagenet_resnext50-32x4d_jpg.py) | 77.604    | 93.856    | 4718    | 12.88    | [model](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/resnext/resnet50/epoch_100.pth) |
+| resnext101-32x4d | [resnext101-32x4d](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/resnext/imagenet_resnext101-32x4d_jpg.py) | 78.568    | 94.344    | 4792    | 26.84    | [model](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/resnext/resnext50-32x4d/epoch_100.pth) |
+| resnext101-32x8d | [resnext101-32x8d](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/resnext/imagenet_resnext101-32x8d_jpg.py) | 79.468    | 94.434    | 9582    | 27.52    | [model](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/resnext/resnext101-32x8d/epoch_100.pth) |
+| resnext152-32x4d | [resnext152-32x4d](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/resnext/imagenet_resnext152-32x4d_jpg.py) | 78.994    | 94.462    | 4852    | 41.08    | [model](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/resnext/resnext152-32x4d/epoch_100.pth) |
+| hrnetw18 | [hrnetw18](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/hrnet/imagenet_hrnetw18_jpg.py) | 76.258    | 92.976    | 4701    | 54.55    | [model](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/hrnet/hrnetw18/epoch_100.pth) |
+| hrnetw30 | [hrnetw30](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/hrnet/imagenet_hrnetw30_jpg.py) | 77.66    | 93.862    | 4766    | 54. 30   | [model](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/hrnet/hrnetw30/epoch_100.pth) |
+| hrnetw32 | [hrnetw32](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/hrnet/imagenet_hrnetw32_jpg.py) | 77.994    | 93.976    | 4780    | 53.48    | [model](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/hrnet/hrnetw32/epoch_100.pth) |
+| hrnetw40 | [hrnetw40](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/hrnet/imagenet_hrnetw40_jpg.py) | 78.142    | 93.956    | 4843    | 54.31    | [model](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/hrnet/hrnetw40/epoch_100.pth) |
+| hrnetw44 | [hrnetw44](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/hrnet/imagenet_hrnetw44_jpg.py) | 79.266    | 94.476    | 4884    | 54.83    | [model](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/hrnet/hrnetw44/epoch_100.pth) |
+| hrnetw48 | [hrnetw48](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/hrnet/imagenet_hrnetw48_jpg.py) | 79.636    | 94.802    | 4916    | 54.14    | [model](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/hrnet/hrnetw48/epoch_100.pth) |
+| hrnetw64 | [hrnetw64](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/hrnet/imagenet_hrnetw64_jpg.py) | 79.884    | 95.04    | 5120    | 54.74    | [model](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/resnet/hrnetw64/epoch_100.pth) |
+| vit-base-patch16 | [vit-base-patch16](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/vit/imagenet_vit_base_patch16_224_jpg.py) | 76.082    | 92.026    | 346    | 8.03    | [model](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/vit/vit-base-patch16/epoch_300.pth) |
+| swin-tiny-patch4-window7 | [swin-tiny-patch4-window7](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/swint/imagenet_swin_tiny_patch4_window7_224_jpg.py) | 80.528    | 94.822    | 132    | 12.94    | [model](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/swint/swin-tiny-patch4-window7/epoch_300.pth) |
+
+(ps: 通过EasyCV训练得到模型结果，推理的输入尺寸默认为224，机器默认为V100 16G，其中gpu memory记录的是gpu peak memory)
+
+| Algorithm | Config                                           | Top-1 (%) | Top-5 (%) | gpu memory (MB) | inference time (ms/img) | Download                                                     |
+| --------- | ------------------------------------------------------------ | --------- | --------- | --------- | --------- | ------------------------------------------------------------ |
+| vit_base_patch16_224 | [vit_base_patch16_224](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/timm/vit/vit_base_patch16_224.py) | 78.096    | 94.324   | 346    | 8.03    | [model](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/timm/vit/B_16-i21k-300ep-lr_0.001-aug_medium1-wd_0.1-do_0.0-sd_0.0--imagenet2012-steps_20k-lr_0.01-res_224.npz) |
+| vit_large_patch16_224 | [vit_large_patch16_224](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/timm/vit/vit_large_patch16_224.py) | 84.404    | 97.276   | 1171    | 16.30    | [model](https://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/timm/vit/L_16-i21k-300ep-lr_0.001-aug_medium1-wd_0.1-do_0.1-sd_0.1--imagenet2012-steps_20k-lr_0.01-res_224.npz) |
+| deit_base_patch16_224 | [deit_base_patch16_224](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/timm/deit/deit_base_patch16_224.py) | 81.756    | 95.6   | 346    | 7.98    | [model](https://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/timm/deit/deit_base_patch16_224-b5f2ef4d.pth) |
+| deit_base_distilled_patch16_224 | [deit_base_distilled_patch16_224](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/timm/deit/deit_base_distilled_patch16_224.py) | 83.232    | 96.476   | 349    | 8.07    | [model](https://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/timm/deit/deit_base_distilled_patch16_224-df68dfff.pth) |
+| xcit_medium_24_p8_224 | [xcit_medium_24_p8_224](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/timm/xcit/xcit_medium_24_p8_224.py) | 83.348    | 96.21   | 884    | 31.77    | [model](https://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/timm/xcit/xcit_medium_24_p8_224.pth) |
+| xcit_medium_24_p8_224_dist | [xcit_medium_24_p8_224_dist](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/timm/xcit/xcit_medium_24_p8_224_dist.py) | 84.876    | 97.164   | 884    | 32.08    | [model](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/timm/xcit/xcit_medium_24_p8_224_dist.pth) |
+| xcit_large_24_p8_224 | [xcit_large_24_p8_224](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/timm/xcit/xcit_large_24_p8_224.py) | 83.986    | 96.47   | 1962    | 37.44    | [model](https://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/timm/xcit/xcit_large_24_p8_224.pth) |
+| xcit_large_24_p8_224_dist | [xcit_large_24_p8_224_dist](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/timm/xcit/xcit_large_24_p8_224_dist.py) | 85.022    | 97.29   | 1962    | 37.44    | [model](https://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/timm/xcit/xcit_large_24_p8_224_dist.pth) |
+| tnt_s_patch16_224 | [tnt_s_patch16_224](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/timm/tnt/tnt_s_patch16_224.py) | 76.934    | 93.388   | 100    | 18.92    | [model](https://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/timm/tnt/tnt_s_patch16_224.pth.tar) |
+| convit_tiny | [convit_tiny](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/timm/convit/convit_tiny.py) | 72.954    | 91.68   | 31    | 10.79    | [model](https://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/timm/convit/convit_tiny.pth) |
+| convit_small | [convit_small](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/timm/convit/convit_small.py) | 81.342   | 95.784   | 122    | 11.23    | [model](https://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/timm/convit/convit_small.pth) |
+| convit_base | [convit_base](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/timm/convit/convit_base.py) | 82.27    | 95.916   | 358    | 11.26    | [model](https://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/timm/convit/convit_base.pth) |
+| cait_xxs24_224 | [cait_xxs24_224](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/timm/cait/cait_xxs24_224.py) | 78.45   | 94.154   | 50    | 22.62    | [model](https://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/timm/cait/XXS24_224.pth) |
+| cait_xxs36_224 | [cait_xxs36_224](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/timm/cait/cait_xxs36_224.py) | 79.788   | 94.87   | 71    | 33.25    | [model](https://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/timm/cait/XXS36_224.pth) |
+| cait_s24_224 | [cait_s24_224](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/timm/cait/cait_s24_224.py) | 83.302   | 96.568   | 190    | 23.74    | [model](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/timm/cait/cait_s24_224.pth) |
+| levit_128 | [levit_128](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/timm/levit/levit_128.py) | 78.468   | 93.874   | 76    | 15.33    | [model](https://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/timm/levit/LeViT-128-b88c2750.pth) |
+| levit_192 | [levit_192](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/timm/levit/levit_192.py) | 79.72   | 94.664   | 128    | 15.17    | [model](https://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/timm/levit/LeViT-192-92712e41.pth) |
+| levit_256 | [levit_256](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/timm/levit/levit_256.py) | 81.432   | 95.38   | 222    | 15.27    | [model](https://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/timm/levit/LeViT-256-13b5763e.pth) |
+| convnext_tiny | [convnext_tiny](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/timm/convnext/convnext_tiny.py) | 81.878   | 95.836   | 128    | 7.17    | [model](https://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/timm/convnext/convnext_tiny_1k_224_ema.pth) |
+| convnext_small | [convnext_small](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/timm/convnext/convnext_small.py) | 82.836   | 96.458   | 213    | 12.89    | [model](https://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/timm/convnext/convnext_small_1k_224_ema.pth) |
+| convnext_base | [convnext_base](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/timm/convnext/convnext_base.py) | 83.73   | 96.692   | 364    | 13.04    | [model](https://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/timm/convnext/convnext_base_1k_224_ema.pth) |
+| convnext_large | [convnext_large](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/timm/convnext/convnext_large.py) | 84.164   | 96.844   | 781    | 13.78    | [model](https://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/timm/convnext/convnext_large_1k_224_ema.pth) |
+| resmlp_12_distilled_224 | [resmlp_12_distilled_224](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/timm/resmlp/resmlp_12_distilled_224.py) | 77.876   | 93.532   | 66    | 4.90    | [model](https://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/timm/resmlp/resmlp_12_dist.pth) |
+| resmlp_24_distilled_224 | [resmlp_24_distilled_224](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/timm/resmlp/resmlp_24_distilled_224.py) | 80.548   | 95.204   | 124    | 9.07    | [model](https://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/timm/resmlp/resmlp_24_dist.pth) |
+| resmlp_36_distilled_224 | [resmlp_36_distilled_224](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/timm/resmlp/resmlp_36_distilled_224.py) | 80.944   | 95.416   | 181    | 13.56    | [model](https://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/timm/resmlp/resmlp_36_dist.pth) |
+| resmlp_big_24_distilled_224 | [resmlp_big_24_distilled_224](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/timm/resmlp/resmlp_big_24_distilled_224.py) | 83.45   | 96.65   | 534    | 20.48    | [model](https://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/timm/resmlp/resmlpB_24_dist.pth) |
+| coat_tiny | [coat_tiny](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/timm/coat/coat_tiny.py) | 78.112   | 93.972   | 127    | 33.09    | [model](https://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/timm/coat/coat_tiny-473c2a20.pth) |
+| coat_mini | [coat_mini](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/timm/coat/coat_mini.py) | 80.912   | 95.378   | 247    | 33.29    | [model](https://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/timm/coat/coat_mini-2c6baf49.pth) |
+| convmixer_768_32 | [convmixer_768_32](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/timm/convmixer/convmixer_768_32.py) | 80.08   | 94.992   | 4995    | 10.23    | [model](https://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/timm/convmixer/convmixer_768_32_ks7_p7_relu.pth.tar) |
+| convmixer_1024_20_ks9_p14 | [convmixer_1024_20_ks9_p14](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/timm/convmixer/convmixer_1024_20_ks9_p14.py) | 81.742   | 95.578   | 2407    | 6.29    | [model](https://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/timm/convmixer/convmixer_1024_20_ks9_p14.pth.tar) |
+| convmixer_1536_20 | [convmixer_1536_20](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/timm/convmixer/convmixer_1536_20.py) | 81.432   | 95.38   | 547    | 14.66    | [model](https://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/timm/convmixer/convmixer_1536_20_ks9_p7.pth.tar) |
+| gmixer_24_224 | [gmixer_24_224](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/timm/gmixer/gmixer_24_224.py) | 78.088   | 93.6   | 104    | 11.65    | [model](https://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/timm/gmixer/gmixer_24_224_raa-7daf7ae6.pth) |
+| gmlp_s16_224 | [gmlp_s16_224](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/timm/gmlp/gmlp_s16_224.py) | 77.204   | 93.358   | 81    | 11.15    | [model](http://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/timm/gmlp/gmlp_s16_224.pth) |
+| mixer_b16_224 | [mixer_b16_224](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/timm/mlp-mixer/mixer_b16_224.py) | 72.558   | 90.068   | 241    | 5.37    | [model](https://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/timm/mlp-mixer/jx_mixer_b16_224-76587d61.pth) |
+| mixer_l16_224 | [mixer_l16_224](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/timm/mlp-mixer/mixer_l16_224.py) | 68.34   | 86.11   | 804    | 11.74    | [model](https://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/timm/mlp-mixer/jx_mixer_l16_224-92f9adc4.pth) |
+| jx_nest_tiny | [jx_nest_tiny](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/timm/nest/jx_nest_tiny.py) | 81.278   | 95.618   | 90    | 9.05    | [model](https://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/timm/nest/jx_nest_tiny-e3428fb9.pth) |
+| jx_nest_small | [jx_nest_tiny](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/timm/nest/jx_nest_small.py) | 83.144   | 96.3   | 174    | 16.92    | [model](https://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/timm/nest/jx_nest_small-422eaded.pth) |
+| jx_nest_base | [jx_nest_base](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/timm/nest/jx_nest_base.py) | 83.474   | 96.442   | 300    | 16.88    | [model](https://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/timm/nest/jx_nest_base-8bc41011.pth) |
+| pit_s_distilled_224 | [pit_s_distilled_224](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/timm/pit/pit_s_distilled_224.py) | 83.144   | 96.3   | 109    | 7.00    | [model](https://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/timm/pit/pit_s_distill_819.pth) |
+| pit_b_distilled_224 | [pit_b_distilled_224](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/timm/pit/pit_b_distilled_224.py) | 83.474   | 96.442   | 330    | 7.66    | [model](https://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/timm/pit/pit_b_distill_840.pth) |
+| twins_svt_small | [twins_svt_small](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/timm/twins/twins_svt_small.py) | 81.598   | 95.55   | 657    | 14.07    | [model](https://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/timm/twins/twins_svt_small-42e5f78c.pth) |
+| twins_svt_base | [twins_svt_base](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/timm/twins/twins_svt_base.py) | 82.882   | 96.234   | 1447    | 18.99    | [model](https://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/timm/twins/twins_svt_base-c2265010.pth) |
+| twins_svt_large | [twins_svt_large](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/timm/twins/twins_svt_large.py) | 83.428   | 96.506   | 2567    | 19.11    | [model](https://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/timm/twins/twins_svt_large-90f6aaa9.pth) |
+| swin_base_patch4_window7_224 | [swin_base_patch4_window7_224](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/timm/swint/swin_base_patch4_window7_224.py) | 84.714   | 97.444   | 375    | 23.47    | [model](https://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/timm/swint/swin_base_patch4_window7_224_22kto1k.pth) |
+| swin_large_patch4_window7_224 | [swin_large_patch4_window7_224](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/timm/swint/swin_large_patch4_window7_224.py) | 85.826   | 97.816   | 788    | 23.29    | [model](https://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/timm/swint/swin_large_patch4_window7_224_22kto1k.pth) |
+| dynamic_swin_small_p4_w7_224 | [dynamic_swin_small_p4_w7_224](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/timm/swint/dynamic_small_base_p4_w7_224.py) | 82.896   | 96.234   | 220    | 28.55    | [model](https://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/timm/swint/swin_small_patch4_window7_224_statedict.pth) |
+| dynamic_swin_tiny_p4_w7_224 | [dynamic_swin_tiny_p4_w7_224](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/timm/swint/dynamic_swin_tiny_p4_w7_224.py) | 80.912   | 95.41   | 136    | 14.58    | [model](https://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/timm/swint/swin_tiny_patch4_window7_224_statedict.pth) |
+| shuffletrans_tiny_p4_w7_224 | [shuffletrans_tiny_p4_w7_224](https://github.com/alibaba/EasyCV/tree/master/configs/classification/imagenet/timm/shuffle_transformer/shuffletrans_tiny_p4_w7_224.py) | 82.176   | 96.05   | 5311    | 13.90    | [model](https://pai-vision-data-hz.oss-cn-zhangjiakou.aliyuncs.com/EasyCV/modelzoo/classification/timm/shuffle_transformer/shuffle_tiny.pth) |
+
+(ps: 通过导入官方模型得到推理结果，需要torch.__version__ >= 1.9.0，推理的输入尺寸默认为224，机器默认为V100 16G，其中gpu memory记录的是gpu peak memory)

easycv/models/backbones/benchmark_mlp.py

@@ -18,11 +18,8 @@ class BenchMarkMLP(nn.Module):
         self.dropout = nn.Dropout(p=0.5)
         self.pool = nn.AdaptiveAvgPool2d((1, 1))
         self.avg_pool = avg_pool
-        # self.fc2 = nn.Linear(feature_num, num_classes)
-        # self.relu2 = nn.ReLU()
-        # self._initialize_weights()
 
-    def init_weights(self, pretrained=None):
+    def init_weights(self):
         for m in self.modules():
             if isinstance(m, nn.Linear):
                 nn.init.kaiming_normal_(
@@ -33,7 +30,4 @@ class BenchMarkMLP(nn.Module):
             x = self.pool(x)
         x = self.fc1(x)
         x = self.relu1(x)
-        # x = self.dropout(x)
-        # x = self.fc2(x)
-        # x = self.relu2(x)
         return tuple([x])

easycv/models/backbones/bninception.py

@@ -344,32 +344,16 @@ class BNInception(nn.Module):
             1024, 128, kernel_size=(1, 1), stride=(1, 1))
         self.inception_5b_pool_proj_bn = nn.BatchNorm2d(128, affine=True)
         self.inception_5b_relu_pool_proj = nn.ReLU(inplace)
-        # self.last_linear = nn.Linear (1024, num_classes)
         self.num_classes = num_classes
         if num_classes > 0:
             self.last_linear = nn.Linear(1024, num_classes)
 
-        self.pretrained = model_urls[self.__class__.__name__]
-
-    def init_weights(self, pretrained=None):
-        if isinstance(pretrained, str) or isinstance(pretrained, dict):
-            logger = get_root_logger()
-            load_checkpoint(self, pretrained, strict=False, logger=logger)
-        elif pretrained is None:
-            for m in self.modules():
-                if isinstance(m, nn.Conv2d):
-                    kaiming_init(m, mode='fan_in', nonlinearity='relu')
-                elif isinstance(m, (_BatchNorm, nn.GroupNorm)):
-                    constant_init(m, 1)
-
-            # if self.zero_init_residual:
-            #     for m in self.modules():
-            #         if isinstance(m, Bottleneck):
-            #             constant_init(m.norm3, 0)
-            #         elif isinstance(m, BasicBlock):
-            #             constant_init(m.norm2, 0)
-        else:
-            raise TypeError('pretrained must be a str or None')
+    def init_weights(self):
+        for m in self.modules():
+            if isinstance(m, nn.Conv2d):
+                kaiming_init(m, mode='fan_in', nonlinearity='relu')
+            elif isinstance(m, (_BatchNorm, nn.GroupNorm)):
+                constant_init(m, 1)
 
     def features(self, input):
         conv1_7x7_s2_out = self.conv1_7x7_s2(input)

easycv/models/backbones/genet.py

@@ -62,7 +62,6 @@ def _fuse_convkx_and_bn_(convkx, bn):
     bn.bias[:] = bn.bias - the_bias_shift
     bn.running_var[:] = 1.0 - bn.eps
     bn.running_mean[:] = 0.0
-    # convkx.register_parameter('bias', bn.bias)
     convkx.bias = nn.Parameter(bn.bias)
 
 
@@ -76,7 +75,6 @@ def remove_bn_in_superblock(super_block):
         for block in the_seq_list:
             if isinstance(block, nn.BatchNorm2d):
                 _fuse_convkx_and_bn_(last_block, block)
-                # print('--debug fuse shortcut bn')
             else:
                 new_seq_list.append(block)
             last_block = block
@@ -92,7 +90,6 @@ def remove_bn_in_superblock(super_block):
         for block in the_seq_list:
             if isinstance(block, nn.BatchNorm2d):
                 _fuse_convkx_and_bn_(last_block, block)
-                # print('--debug fuse conv bn')
             else:
                 new_seq_list.append(block)
             last_block = block
@@ -1647,7 +1644,6 @@ class PlainNet(nn.Module):
             block_list.pop(-1)
         else:
             self.adptive_avg_pool = nn.AdaptiveAvgPool2d((1, 1))
-            # AdaptiveAvgPool(out_channels=block_list[-1].out_channels, output_size=1)
 
         self.block_list = block_list
         if not no_create:
@@ -1663,29 +1659,13 @@ class PlainNet(nn.Module):
 
         self.plainnet_struct = str(self) + str(self.adptive_avg_pool)
         self.zero_init_residual = False
-        self.pretrained = model_urls[self.__class__.__name__ +
-                                     plainnet_struct_idx]
 
-    def init_weights(self, pretrained=None):
-        if isinstance(pretrained, str) or isinstance(pretrained, dict):
-            logger = get_root_logger()
-            load_checkpoint(self, pretrained, strict=False, logger=logger)
-        elif pretrained is None:
-            for m in self.modules():
-                if isinstance(m, nn.Conv2d):
-                    kaiming_init(m, mode='fan_in', nonlinearity='relu')
-                elif isinstance(m, (_BatchNorm, nn.GroupNorm)):
-                    constant_init(m, 1)
-
-            # if self.zero_init_residual:
-            #     for m in self.modules():
-            #         if isinstance(m, Bottleneck):
-            #             constant_init(m.norm3, 0)
-            #         elif isinstance(m, BasicBlock):
-            #             constant_init(m.norm2, 0)
-        else:
-            raise TypeError('pretrained must be a str or None')
-        return
+    def init_weights(self):
+        for m in self.modules():
+            if isinstance(m, nn.Conv2d):
+                kaiming_init(m, mode='fan_in', nonlinearity='relu')
+            elif isinstance(m, (_BatchNorm, nn.GroupNorm)):
+                constant_init(m, 1)
 
     def forward(self, x):
         output = x
@@ -1699,13 +1679,3 @@ class PlainNet(nn.Module):
             output = self.fc_linear(output)
 
         return [output]
-
-
-if __name__ == '__main__':
-    from torch import nn
-
-    input_image_size = 192
-
-    # model = genet_normal(pretrained=True, root='data/models/GENet_params/')
-    # print(type(model))
-    # print(model.block_list)

easycv/models/backbones/hrnet.py

@@ -184,7 +184,6 @@ class Bottleneck(nn.Module):
 
         if self.with_cp and x.requires_grad:
             raise NotImplementedError
-            # out = cp.checkpoint(_inner_forward, x)
         else:
             out = _inner_forward(x)
 
@@ -718,31 +717,25 @@ class HRNet(nn.Module):
 
         return nn.Sequential(*hr_modules), in_channels
 
-    def init_weights(self, pretrained=None):
+    def init_weights(self):
         """Initialize the weights in backbone.
 
         Args:
             pretrained (str, optional): Path to pre-trained weights.
                 Defaults to None.
         """
-        if isinstance(pretrained, str):
-            logger = get_root_logger()
-            load_checkpoint(self, pretrained, strict=False, logger=logger)
-        elif pretrained is None:
-            for m in self.modules():
-                if isinstance(m, nn.Conv2d):
-                    normal_init(m, std=0.001)
-                elif isinstance(m, (_BatchNorm, nn.GroupNorm)):
-                    constant_init(m, 1)
+        for m in self.modules():
+            if isinstance(m, nn.Conv2d):
+                normal_init(m, std=0.001)
+            elif isinstance(m, (_BatchNorm, nn.GroupNorm)):
+                constant_init(m, 1)
 
-            if self.zero_init_residual:
-                for m in self.modules():
-                    if isinstance(m, Bottleneck):
-                        constant_init(m.norm3, 0)
-                    elif isinstance(m, BasicBlock):
-                        constant_init(m.norm2, 0)
-        else:
-            raise TypeError('pretrained must be a str or None')
+        if self.zero_init_residual:
+            for m in self.modules():
+                if isinstance(m, Bottleneck):
+                    constant_init(m.norm3, 0)
+                elif isinstance(m, BasicBlock):
+                    constant_init(m.norm2, 0)
 
     def forward(self, x):
         """Forward function."""

easycv/models/backbones/inceptionv3.py

@@ -60,23 +60,17 @@ class Inception3(nn.Module):
         if num_classes > 0:
             self.fc = nn.Linear(2048, num_classes)
 
-        self.pretrained = model_urls[self.__class__.__name__]
-
-    def init_weights(self, pretrained=None):
-        if pretrained is not None:
-            logger = get_root_logger()
-            load_checkpoint(self, pretrained, strict=False, logger=logger)
-        else:
-            for m in self.modules():
-                if isinstance(m, nn.Conv2d) or isinstance(m, nn.Linear):
-                    for m in self.modules():
-                        if isinstance(m, nn.Conv2d):
-                            kaiming_init(m, mode='fan_in', nonlinearity='relu')
-                        elif isinstance(m, (_BatchNorm, nn.GroupNorm)):
-                            constant_init(m, 1)
-                elif isinstance(m, nn.BatchNorm2d):
-                    nn.init.constant_(m.weight, 1)
-                    nn.init.constant_(m.bias, 0)
+    def init_weights(self):
+        for m in self.modules():
+            if isinstance(m, nn.Conv2d) or isinstance(m, nn.Linear):
+                for m in self.modules():
+                    if isinstance(m, nn.Conv2d):
+                        kaiming_init(m, mode='fan_in', nonlinearity='relu')
+                    elif isinstance(m, (_BatchNorm, nn.GroupNorm)):
+                        constant_init(m, 1)
+            elif isinstance(m, nn.BatchNorm2d):
+                nn.init.constant_(m.weight, 1)
+                nn.init.constant_(m.bias, 0)
 
     def forward(self, x):
         if self.transform_input:

easycv/models/backbones/lighthrnet.py

@@ -958,24 +958,18 @@ class LiteHRNet(nn.Module):
 
         return nn.Sequential(*modules), in_channels
 
-    def init_weights(self, pretrained=None):
+    def init_weights(self):
         """Initialize the weights in backbone.
 
         Args:
             pretrained (str, optional): Path to pre-trained weights.
                 Defaults to None.
         """
-        if isinstance(pretrained, str):
-            logger = get_root_logger()
-            load_checkpoint(self, pretrained, strict=False, logger=logger)
-        elif pretrained is None:
-            for m in self.modules():
-                if isinstance(m, nn.Conv2d):
-                    normal_init(m, std=0.001)
-                elif isinstance(m, (_BatchNorm, nn.GroupNorm)):
-                    constant_init(m, 1)
-        else:
-            raise TypeError('pretrained must be a str or None')
+        for m in self.modules():
+            if isinstance(m, nn.Conv2d):
+                normal_init(m, std=0.001)
+            elif isinstance(m, (_BatchNorm, nn.GroupNorm)):
+                constant_init(m, 1)
 
     def forward(self, x):
         """Forward function."""

easycv/models/backbones/mae_vit_transformer.py

@@ -73,17 +73,16 @@ class MaskedAutoencoderViT(nn.Module):
         ])
         self.norm = norm_layer(embed_dim)
 
-        self.apply(self._init_weights)
-
-    def _init_weights(self, m):
-        if isinstance(m, nn.Linear):
-            # we use xavier_uniform following official JAX ViT:
-            torch.nn.init.xavier_uniform_(m.weight)
-            if isinstance(m, nn.Linear) and m.bias is not None:
+    def init_weights(self):
+        for m in self.modules():
+            if isinstance(m, nn.Linear):
+                # we use xavier_uniform following official JAX ViT:
+                torch.nn.init.xavier_uniform_(m.weight)
+                if isinstance(m, nn.Linear) and m.bias is not None:
+                    nn.init.constant_(m.bias, 0)
+            elif isinstance(m, nn.LayerNorm):
                 nn.init.constant_(m.bias, 0)
-        elif isinstance(m, nn.LayerNorm):
-            nn.init.constant_(m.bias, 0)
-            nn.init.constant_(m.weight, 1.0)
+                nn.init.constant_(m.weight, 1.0)
 
     def random_masking(self, x, mask_ratio):
         """

easycv/models/backbones/mnasnet.py

@@ -143,8 +143,6 @@ class MNASNet(torch.nn.Module):
             self.classifier = nn.Sequential(
                 nn.Dropout(p=dropout, inplace=True),
                 nn.Linear(1280, num_classes))
-        self.init_weights()
-        self.pretrained = model_urls[self.__class__.__name__ + str(alpha)]
 
     def forward(self, x):
         x = self.layers(x)
@@ -155,22 +153,16 @@ class MNASNet(torch.nn.Module):
         else:
             return [x]
 
-    def init_weights(self, pretrained=None):
-        if isinstance(pretrained, str) or isinstance(pretrained, dict):
-            logger = get_root_logger()
-            load_checkpoint(self, pretrained, strict=False, logger=logger)
-        elif pretrained is None:
-            for m in self.modules():
-                if isinstance(m, nn.Conv2d):
-                    nn.init.kaiming_normal_(
-                        m.weight, mode='fan_out', nonlinearity='relu')
-                    if m.bias is not None:
-                        nn.init.zeros_(m.bias)
-                elif isinstance(m, nn.BatchNorm2d):
-                    nn.init.ones_(m.weight)
+    def init_weights(self):
+        for m in self.modules():
+            if isinstance(m, nn.Conv2d):
+                nn.init.kaiming_normal_(
+                    m.weight, mode='fan_out', nonlinearity='relu')
+                if m.bias is not None:
                     nn.init.zeros_(m.bias)
-                elif isinstance(m, nn.Linear):
-                    nn.init.normal_(m.weight, 0.01)
-                    nn.init.zeros_(m.bias)
-        else:
-            raise TypeError('pretrained must be a str or None')
+            elif isinstance(m, nn.BatchNorm2d):
+                nn.init.ones_(m.weight)
+                nn.init.zeros_(m.bias)
+            elif isinstance(m, nn.Linear):
+                nn.init.normal_(m.weight, 0.01)
+                nn.init.zeros_(m.bias)

easycv/models/backbones/mobilenetv2.py

@@ -156,24 +156,18 @@ class MobileNetV2(nn.Module):
         self.pretrained = model_urls[self.__class__.__name__ + '_' +
                                      str(width_multi)]
 
-    def init_weights(self, pretrained=None):
-        if isinstance(pretrained, str) or isinstance(pretrained, dict):
-            logger = get_root_logger()
-            load_checkpoint(self, pretrained, strict=False, logger=logger)
-        elif pretrained is None:
-            for m in self.modules():
-                if isinstance(m, nn.Conv2d):
-                    nn.init.kaiming_normal_(m.weight, mode='fan_out')
-                    if m.bias is not None:
-                        nn.init.zeros_(m.bias)
-                elif isinstance(m, nn.BatchNorm2d):
-                    nn.init.ones_(m.weight)
+    def init_weights(self):
+        for m in self.modules():
+            if isinstance(m, nn.Conv2d):
+                nn.init.kaiming_normal_(m.weight, mode='fan_out')
+                if m.bias is not None:
                     nn.init.zeros_(m.bias)
-                elif isinstance(m, nn.Linear):
-                    nn.init.normal_(m.weight, 0, 0.01)
-                    nn.init.zeros_(m.bias)
-        else:
-            raise TypeError('pretrained must be a str or None')
+            elif isinstance(m, nn.BatchNorm2d):
+                nn.init.ones_(m.weight)
+                nn.init.zeros_(m.bias)
+            elif isinstance(m, nn.Linear):
+                nn.init.normal_(m.weight, 0, 0.01)
+                nn.init.zeros_(m.bias)
 
     def forward(self, x):
         x = self.features(x)

easycv/models/backbones/pytorch_image_models_wrapper.py

@@ -1,12 +1,15 @@
 # Copyright (c) Alibaba, Inc. and its affiliates.
+import importlib
 from distutils.version import LooseVersion
 
 import timm
 import torch
 import torch.nn as nn
+from timm.models.helpers import load_pretrained
+from timm.models.hub import download_cached_file
 
 from easycv.utils.checkpoint import load_checkpoint
-from easycv.utils.logger import get_root_logger
+from easycv.utils.logger import get_root_logger, print_log
 from ..modelzoo import timm_models as model_urls
 from ..registry import BACKBONES
 from .shuffle_transformer import (shuffletrans_base_p4_w7_224,
@@ -66,8 +69,6 @@ class PytorchImageModelWrapper(nn.Module):
 
     def __init__(self,
                  model_name='resnet50',
-                 pretrained=False,
-                 checkpoint_path=None,
                  scriptable=None,
                  exportable=None,
                  no_jit=None,
@@ -76,15 +77,16 @@ class PytorchImageModelWrapper(nn.Module):
         Inits PytorchImageModelWrapper by timm.create_models
         Args:
             model_name (str): name of model to instantiate
-            pretrained (bool): load pretrained ImageNet-1k weights if true
-            checkpoint_path (str): path of checkpoint to load after model is initialized
             scriptable (bool): set layer config so that model is jit scriptable (not working for all models yet)
             exportable (bool): set layer config so that model is traceable / ONNX exportable (not fully impl/obeyed yet)
             no_jit (bool): set layer config so that model doesn't utilize jit scripted layers (so far activations only)
         """
         super(PytorchImageModelWrapper, self).__init__()
 
+        self.model_name = model_name
+
         timm_model_names = timm.list_models(pretrained=False)
+        self.timm_model_names = timm_model_names
         assert model_name in timm_model_names or model_name in _MODEL_MAP, \
             f'{model_name} is not in model_list of timm/fair, please check the model_name!'
 
@@ -94,52 +96,52 @@ class PytorchImageModelWrapper(nn.Module):
 
         # create model by timm
         if model_name in timm_model_names:
-            try:
-                if pretrained and (model_name in model_urls):
-                    self.model = timm.create_model(model_name, False, '',
-                                                   scriptable, exportable,
-                                                   no_jit, **kwargs)
-                    self.init_weights(model_urls[model_name])
-                    print('Info: Load model from %s' % model_urls[model_name])
-
-                    if checkpoint_path is not None:
-                        self.init_weights(checkpoint_path)
-                else:
-                    # load from timm
-                    if pretrained and model_name.startswith('swin_') and (
-                            LooseVersion(
-                                torch.__version__) <= LooseVersion('1.6.0')):
-                        print(
-                            'Warning: Pretrained SwinTransformer from timm may be zipfile extract'
-                            ' error while torch<=1.6.0')
-                    self.model = timm.create_model(model_name, pretrained,
-                                                   checkpoint_path, scriptable,
-                                                   exportable, no_jit,
-                                                   **kwargs)
-
-            # need fix: delete this except after pytorch 1.7 update in all production
-            # (dlc, dsw, studio, ev_predict_py3)
-            except Exception:
-                print(
-                    f'Error: Fail to create {model_name} with (pretrained={pretrained}, checkpoint_path={checkpoint_path} ...)'
-                )
-                print(
-                    f'Try to create {model_name} with pretrained=False, checkpoint_path=None and default params'
-                )
-                self.model = timm.create_model(model_name, False, '', None,
-                                               None, None, **kwargs)
-
-        # facebook model wrapper
-        if model_name in _MODEL_MAP:
+            self.model = timm.create_model(model_name, False, '', scriptable,
+                                           exportable, no_jit, **kwargs)
+        elif model_name in _MODEL_MAP:
             self.model = _MODEL_MAP[model_name](**kwargs)
-            if pretrained:
-                if model_name in model_urls.keys():
+
+    def init_weights(self, pretrained=None):
+        """
+        Args:
+            if pretrained == True, load model from default path;
+            if pretrained == False or None, load from init weights.
+
+            if model_name in timm_model_names, load model from timm default path;
+            if model_name in _MODEL_MAP, load model from easycv default path
+        """
+        logger = get_root_logger()
+        if pretrained:
+            if self.model_name in self.timm_model_names:
+                pretrained_path = model_urls[self.model_name]
+                print_log(
+                    'load model from default path: {}'.format(pretrained_path),
+                    logger)
+                if pretrained_path.endswith('.npz'):
+                    pretrained_loc = download_cached_file(
+                        pretrained_path, check_hash=False, progress=False)
+                    return self.model.load_pretrained(pretrained_loc)
+                else:
+                    backbone_module = importlib.import_module(
+                        self.model.__module__)
+                    return load_pretrained(
+                        self.model,
+                        default_cfg={'url': pretrained_path},
+                        filter_fn=backbone_module.checkpoint_filter_fn
+                        if hasattr(backbone_module, 'checkpoint_filter_fn')
+                        else None)
+            elif self.model_name in _MODEL_MAP:
+                if self.model_name in model_urls.keys():
+                    pretrained_path = model_urls[self.model_name]
+                    print_log(
+                        'load model from default path: {}'.format(
+                            pretrained_path), logger)
                     try_max = 3
                     try_idx = 0
                     while try_idx < try_max:
                         try:
                             state_dict = torch.hub.load_state_dict_from_url(
-                                url=model_urls[model_name],
+                                url=pretrained_path,
                                 map_location='cpu',
                             )
                             try_idx += try_max
@@ -147,31 +149,22 @@ class PytorchImageModelWrapper(nn.Module):
                             try_idx += 1
                             state_dict = {}
                             if try_idx == try_max:
-                                print(
-                                    'load from url failed ! oh my DLC & OSS, you boys really good! ',
-                                    model_urls[model_name])
+                                print_log(
+                                    f'load from url failed ! oh my DLC & OSS, you boys really good! {model_urls[self.model_name]}',
+                                    logger)
 
-                    # for some model strict = False still failed when model doesn't exactly match
-                    try:
-                        self.model.load_state_dict(state_dict, strict=False)
-                    except Exception:
-                        print('load for model_name not all right')
+                    if 'model' in state_dict:
+                        state_dict = state_dict['model']
+                    self.model.load_state_dict(state_dict, strict=False)
                 else:
-                    print('%s not in evtorch modelzoo!' % model_name)
-
-    def init_weights(self, pretrained=None):
-        # pretrained is the path of pretrained model offered by easycv
-        if pretrained is not None:
-            logger = get_root_logger()
-            load_checkpoint(
-                self.model,
-                pretrained,
-                map_location=torch.device('cpu'),
-                strict=False,
-                logger=logger)
+                    raise ValueError('{} not in evtorch modelzoo!'.format(
+                        self.model_name))
+            else:
+                raise ValueError(
+                    'Error: Fail to create {} with (pretrained={}...)'.format(
+                        self.model_name, pretrained))
         else:
-            # init by timm
-            pass
+            self.model.init_weights()
 
     def forward(self, x):
 

easycv/models/backbones/resnest.py

@@ -470,26 +470,19 @@ class ResNeSt(nn.Module):
         self.avgpool = GlobalAvgPool2d()
         self.drop = nn.Dropout(final_drop) if final_drop > 0.0 else None
         self.norm_layer = norm_layer
-        # self.fc = nn.Linear(512 * block.expansion, num_classes)
 
         if num_classes > 0:
             self.avgpool = nn.AdaptiveAvgPool2d((1, 1))
             self.fc = nn.Linear(512 * block.expansion, num_classes)
 
-    def init_weights(self, pretrained=None):
-        if isinstance(pretrained, str) or isinstance(pretrained, dict):
-            logger = get_root_logger()
-            load_checkpoint(self, pretrained, strict=False, logger=logger)
-        elif pretrained is None:
-            for m in self.modules():
-                if isinstance(m, nn.Conv2d):
-                    n = m.kernel_size[0] * m.kernel_size[1] * m.out_channels
-                    m.weight.data.normal_(0, math.sqrt(2. / n))
-                elif isinstance(m, self.norm_layer):
-                    m.weight.data.fill_(1)
-                    m.bias.data.zero_()
-        else:
-            raise TypeError('pretrained must be a str or None')
+    def init_weights(self):
+        for m in self.modules():
+            if isinstance(m, nn.Conv2d):
+                n = m.kernel_size[0] * m.kernel_size[1] * m.out_channels
+                m.weight.data.normal_(0, math.sqrt(2. / n))
+            elif isinstance(m, self.norm_layer):
+                m.weight.data.fill_(1)
+                m.bias.data.zero_()
 
     def _make_layer(self,
                     block,
@@ -613,7 +606,6 @@ class ResNeSt(nn.Module):
 
         if hasattr(self, 'fc'):
             x = self.avgpool(x)
-            # x = x.view(x.size(0), -1)
             x = torch.flatten(x, 1)
             if self.drop:
                 x = self.drop(x)