06 Pytorch Transfer Learning -> Setting up a pretrained model

[avishekiitism]

My error

RuntimeError: invalid hash value (expected "3dd342df", got "7f5810bc96def8f7552d5b7e68d53c4786f81167d28291b21c0d90e1fca14934")

My code

# OLD: Setup the model with pretrained weights and send it to the target device (this was prior to torchvision v0.13)
# model = torchvision.models.efficientnet_b0(pretrained=True).to(device) # OLD method (with pretrained=True)

# NEW: Setup the model with pretrained weights and send it to the target device (torchvision v0.13+)
weights = torchvision.models.EfficientNet_B0_Weights.DEFAULT # .DEFAULT = best available weights 
model = torchvision.models.efficientnet_b0(weights=weights).to(device)

#model # uncomment to output (it's very long)

[premsai112]

NEW: Setup the model with pre-trained weights and send it to the target device (torchvision v0.13+) Hi I am encountering the same error. Can someone help me with this?

[premsai112]

"RuntimeError Traceback (most recent call last)

in <cell line: 6>()
4 # NEW: Setup the model with pretrained weights and send it to the target device (torchvision v0.13+)
5 weights = torchvision.models.EfficientNet_B0_Weights.DEFAULT # .DEFAULT = best available weights
----> 6 model = torchvision.models.efficientnet_b0(weights=weights).to(device)
7
8 model # uncomment to output (it's very long)

6 frames

/usr/local/lib/python3.10/dist-packages/torch/hub.py in download_url_to_file(url, dst, hash_prefix, progress)
661 digest = sha256.hexdigest()
662 if digest[:len(hash_prefix)] != hash_prefix:
--> 663 raise RuntimeError(f'invalid hash value (expected "{hash_prefix}", got "{digest}")')
664 shutil.move(f.name, dst)
665 finally:

RuntimeError: invalid hash value (expected "3dd342df", got "7f5810bc96def8f7552d5b7e68d53c4786f81167d28291b21c0d90e1fca14934")"

[premsai112]

@mrdbourke can you please help with this output error?

[DarekW90]

Hey, I have same problem, but according to PyTorch repo this bug has been reported and its going to be fixed soon.
(pytorch/vision#7744 (comment))

To bypass this bug for now you should simply run this code one cell above :)

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")

Hope its helps :)

Cheers!

[premsai112]

Thanks man! *Premsai Karampudi**, MBA, RIPC*

…

On Mon, 29 Jan 2024 at 6:52 PM, DarekW90 ***@***.***> wrote: Hey, I have same problem, but according to PyTorch repo this bug has been reported and its going to be fixed soon. (pytorch/vision#7744 (comment) <pytorch/vision#7744 (comment)>) To bypass this bug for now you should simply run this code: 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") One cell above :) hope its helps :) Cheers! — Reply to this email directly, view it on GitHub <#771 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AWXSCR7YDQQRSRXP2M7O4FLYQ6PBHAVCNFSM6AAAAABBBLBIBGVHI2DSMVQWIX3LMV43SRDJONRXK43TNFXW4Q3PNVWWK3TUHM4DEOBSGYYDS> . You are receiving this because you commented.Message ID: <mrdbourke/pytorch-deep-learning/repo-discussions/771/comments/8282609@ github.com>