mirrors/pytorch-image-models
1
0
Fork 0
mirror of https://github.com/huggingface/pytorch-image-models.git synced 2025-06-03 15:01:08 +08:00
Code Issues Projects Releases Wiki Activity

Add 34/34d pre-act resnet variants

Browse Source
This commit is contained in:
Ross Wightman 2024-10-14 13:23:50 -07:00
parent 72f0edb7e8
commit abdf33145c
1 changed files with 22 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

22
timm/models/resnetv2.py
View File

@ -700,6 +700,10 @@ default_cfgs = generate_default_cfgs({
interpolation='bicubic', crop_pct=0.95), interpolation='bicubic', crop_pct=0.95),
'resnetv2_18d.untrained': _cfg( 'resnetv2_18d.untrained': _cfg(
interpolation='bicubic', crop_pct=0.95, first_conv='stem.conv1'), interpolation='bicubic', crop_pct=0.95, first_conv='stem.conv1'),
'resnetv2_34.untrained': _cfg(
interpolation='bicubic', crop_pct=0.95),
'resnetv2_34d.untrained': _cfg(
interpolation='bicubic', crop_pct=0.95, first_conv='stem.conv1'),
'resnetv2_50.a1h_in1k': _cfg( 'resnetv2_50.a1h_in1k': _cfg(
hf_hub_id='timm/', hf_hub_id='timm/',
interpolation='bicubic', crop_pct=0.95, test_input_size=(3, 288, 288), test_crop_pct=1.0), interpolation='bicubic', crop_pct=0.95, test_input_size=(3, 288, 288), test_crop_pct=1.0),
@ -784,6 +788,24 @@ def resnetv2_18d(pretrained=False, **kwargs) -> ResNetV2:
return _create_resnetv2('resnetv2_18d', pretrained=pretrained, **dict(model_args, **kwargs)) return _create_resnetv2('resnetv2_18d', pretrained=pretrained, **dict(model_args, **kwargs))
@register_model
def resnetv2_34(pretrained=False, **kwargs) -> ResNetV2:
model_args = dict(
layers=(3, 4, 6, 3), channels=(64, 128, 256, 512), basic=True, bottle_ratio=1.0,
conv_layer=create_conv2d, norm_layer=BatchNormAct2d
)
return _create_resnetv2('resnetv2_34', pretrained=pretrained, **dict(model_args, **kwargs))
@register_model
def resnetv2_34d(pretrained=False, **kwargs) -> ResNetV2:
model_args = dict(
layers=(3, 4, 6, 3), channels=(64, 128, 256, 512), basic=True, bottle_ratio=1.0,
conv_layer=create_conv2d, norm_layer=BatchNormAct2d, stem_type='deep', avg_down=True
)
return _create_resnetv2('resnetv2_34d', pretrained=pretrained, **dict(model_args, **kwargs))
@register_model @register_model
def resnetv2_50(pretrained=False, **kwargs) -> ResNetV2: def resnetv2_50(pretrained=False, **kwargs) -> ResNetV2:
model_args = dict(layers=[3, 4, 6, 3], conv_layer=create_conv2d, norm_layer=BatchNormAct2d) model_args = dict(layers=[3, 4, 6, 3], conv_layer=create_conv2d, norm_layer=BatchNormAct2d)