How to solve this problem?

[OhJackHu]

Hi, bro, your code is very beautiful, but I made the following error when using swin-t to run the code. How can I use the pre training weight downloaded by myself instead of downloading at run time.
Traceback (most recent call last):
File "train.py", line 396, in
train_loader, test_loader, model, optimizer, schedule = set_environment(args)
File "train.py", line 94, in set_environment
model = SwinVit12(
File "/home/pengtl/jackhu/FGVC-PIM-master/models/SwinVit12.py", line 202, in init
self.extractor = timm.create_model('swin_large_patch4_window12_384_in22k', pretrained=True)
File "/home/pengtl/jackhu/FGVC-PIM-master/timm/models/factory.py", line 81, in create_model
model = create_fn(pretrained=pretrained, **kwargs)
File "/home/pengtl/jackhu/FGVC-PIM-master/timm/models/swin_transformer.py", line 654, in swin_large_patch4_window12_384_in22k
model = _create_swin_transformer('swin_large_patch4_window12_384_in22k', pretrained=pretrained, **model_kwargs)
File "/home/pengtl/jackhu/FGVC-PIM-master/timm/models/swin_transformer.py", line 562, in _create_swin_transformer
model = build_model_with_cfg(
File "/home/pengtl/jackhu/FGVC-PIM-master/timm/models/helpers.py", line 457, in build_model_with_cfg
load_pretrained(
File "/home/pengtl/jackhu/FGVC-PIM-master/timm/models/helpers.py", line 184, in load_pretrained
state_dict = load_state_dict_from_url(pretrained_url, progress=progress, map_location='cpu')
File "/home/pengtl/anaconda3/lib/python3.8/site-packages/torch/hub.py", line 528, in load_state_dict_from_url
return torch.load(cached_file, map_location=map_location)
File "/home/pengtl/anaconda3/lib/python3.8/site-packages/torch/serialization.py", line 593, in load
return _legacy_load(opened_file, map_location, pickle_module, **pickle_load_args)
File "/home/pengtl/anaconda3/lib/python3.8/site-packages/torch/serialization.py", line 762, in _legacy_load
magic_number = pickle_module.load(f, **pickle_load_args)
_pickle.UnpicklingError: invalid load key, '<'.

[OhJackHu]

If I use the following code, can I avoid downloading pre training weights？

self.extractor = timm.create_model('swin_large_patch4_window12_384_in22k', pretrained=False)
self.extractor = load_model_weights(self.extractor, './swin_large_patch4_window12_384_in22k.pth')

[chou141253]

Sure! But we use timm model, please check the pre-trained model is fitting timm provided.

[OhJackHu]

I used the pre-training weight downloaded by myself, and the following errors occurred:

[zzr9zzr]

I used the pre-training weight downloaded by myself, and the following errors occurred:

you should change the 'state_dict' to 'model'.