|
|
|
@ -422,7 +422,8 @@ def resmlp_24_224(pretrained=False, **kwargs):
|
|
|
|
|
Paper: `ResMLP: Feedforward networks for image classification...` - https://arxiv.org/abs/2105.03404
|
|
|
|
|
"""
|
|
|
|
|
model_args = dict(
|
|
|
|
|
patch_size=16, num_blocks=24, hidden_dim=384, mlp_ratio=4, block_layer=ResBlock, norm_layer=Affine, **kwargs)
|
|
|
|
|
patch_size=16, num_blocks=24, hidden_dim=384, mlp_ratio=4,
|
|
|
|
|
block_layer=partial(ResBlock, init_values=1e-5), norm_layer=Affine, **kwargs)
|
|
|
|
|
model = _create_mixer('resmlp_24_224', pretrained=pretrained, **model_args)
|
|
|
|
|
return model
|
|
|
|
|
|
|
|
|
@ -433,7 +434,8 @@ def resmlp_36_224(pretrained=False, **kwargs):
|
|
|
|
|
Paper: `ResMLP: Feedforward networks for image classification...` - https://arxiv.org/abs/2105.03404
|
|
|
|
|
"""
|
|
|
|
|
model_args = dict(
|
|
|
|
|
patch_size=16, num_blocks=36, hidden_dim=384, mlp_ratio=4, block_layer=ResBlock, norm_layer=Affine, **kwargs)
|
|
|
|
|
patch_size=16, num_blocks=36, hidden_dim=384, mlp_ratio=4,
|
|
|
|
|
block_layer=partial(ResBlock, init_values=1e-5), norm_layer=Affine, **kwargs)
|
|
|
|
|
model = _create_mixer('resmlp_36_224', pretrained=pretrained, **model_args)
|
|
|
|
|
return model
|
|
|
|
|
|
|
|
|
|