Video 281 - Creating an EffNetB2 Feature Extractor Model - invalid hash

[MariaKrakowiak]

Hi All,

I have come across a problem with below fragment of code:

I have received below:

Did anyone have similar troubles? If so - could you please advise?

Thanks in advance :)

[gulnuravci]

Hi @MariaKrakowiak, I just had the same issue. Here is the fix I found that worked for me: pytorch/vision#7744 Hope this helps!

[gulnuravci]

from torchvision.models import efficientnet_b0, EfficientNet_B0_Weights
from torchvision.models._api import WeightsEnum
from torch.hub import load_state_dict_from_url

def get_state_dict(self, *args, **kwargs):
    kwargs.pop("check_hash")
    return load_state_dict_from_url(self.url, *args, **kwargs)
WeightsEnum.get_state_dict = get_state_dict

efficientnet_b0(weights=EfficientNet_B0_Weights.IMAGENET1K_V1)
efficientnet_b0(weights="DEFAULT")

[gulnuravci]

Specifically this ^

[MariaKrakowiak]

Hi @gulnuravci - your suggestion works. Many thanks for your advice - much appreciated 👍

[mrdbourke]

Hi @gulnuravci ,

Thank you for the solution! I didn't know about this.

Also for those wondering, see the link to the solution comment here: pytorch/vision#7744 (comment)

From the looks of that thread^ it seems like a PyTorch torchvision issue that should be fixed in a future version.

[gulnuravci]

@MariaKrakowiak I'm glad to hear that and @mrdbourke I'm glad to have helped, thank you for all your efforts, it has been inspiring!