safetensors_rust.SafetensorError: Error while deserializing header: HeaderTooLarge

[meongeun]

An issue arises in the code at line 3284 of modeling_utils.py, specifically with the following snippet:

with safe_open(resolved_archive_file, framework="pt") as f

Have you encountered this error before, and do you have any insights into its cause?

[devin12422]

I'm getting the same error, quick google search say's it's an issue with the archive the weights are stored in.

[Liuhm0710]

Traceback (most recent call last): File "/./diffumask/myenv/lib/python3.8/site-packages/diffusers/models/modeling_utils.py", line 102, in load_state_dict return safetensors.torch.load_file(checkpoint_file, device="cpu") File "/./diffumask/myenv/lib/python3.8/site-packages/safetensors/torch.py", line 259, in load_file with safe_open(filename, framework="pt", device=device) as f: safetensors_rust.SafetensorError: Error while deserializing header: HeaderTooLarge

facing the same problem too

[billl-jiang]

Hello,

The SafetensorError: Error while deserializing header: HeaderTooLarge you've encountered is similar to issue #68, often due to issues with downloading checkpoints. To address this, please use Git LFS to download the necessary large files:

git lfs pull --include="<files to download>" --exclude="<files to exclude>"

Replace <files to download> and <files to exclude> as needed. This should download the actual file content, resolving the error.

[quietscientist]

edit: this worked, the issue was an incorrectly downloaded flan-t5 checkpoint, not the motiongpt checkpoint. Thanks!

[cartucho-move]

sudo apt-get install git-lfs may also be needed if on Ubuntu