[BUG] Fail to turn swin_base_patch4_window12_384 into feature extractor

[rentainhe]

Hi, I'm trying to use swin_base_patch4_window12_384 model to extract features, but I meet some error as follows:

model = timm.create_model("swin_base_patch4_window12_384", features_only=True, pretrained=False)

AttributeError: 'SwinTransformer' object has no attribute 'feature_info'

it seems like that there's no feature_info in SwinTransformer

[Bailey-24]

Hi, do you have extracted features from swin v2?

[rwightman]

@rentainhe @Bailey-24 took some time to get around to this, but on main branch, all swin v1 & v2 models support feat extraction now. Be aware they are NHWC outputs, not NCHW like most convnets. EDIT on main branch right now, in pypi pre-release in days and hopefully a non pre-release cycle happens in a month or so (0.9).

model = timm.create_model("swin_base_patch4_window12_384", features_only=True, pretrained=False)
oo = model(torch.randn(2, 3, 384, 384))
for x in oo:
    print(x.shape)

torch.Size([2, 96, 96, 128])
torch.Size([2, 48, 48, 256])
torch.Size([2, 24, 24, 512])
torch.Size([2, 12, 12, 1024])

[Mojzaar]

When I try your code I get this: AttributeError: 'SwinTransformer' object has no attribute 'feature_info'

[rwightman]

Old version, need recent 0.8.x see readme

On Thu, Apr 6, 2023, 10:28 AM Mojtaba @.***> wrote:

When I try your code I get this: AttributeError: 'SwinTransformer' object has no attribute 'feature_info'

— Reply to this email directly, view it on GitHub https://github.com/huggingface/pytorch-image-models/issues/1438#issuecomment-1499395359, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABLQICASBAIXU6AQOLCIULDW734KVANCNFSM57WC6D2Q . You are receiving this because you modified the open/close state.Message ID: @.***>