Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Fix two torch warnings #34

Open
wants to merge 2 commits into
base: main
Choose a base branch
from

Conversation

DougHaber
Copy link

When using this module, two warnings are displayed:

./python3.12/site-packages/RealESRGAN/model.py:59: FutureWarning: torch.cuda.amp.autocast(args...) is deprecated. Please use torch.amp.autocast('cuda', args...) instead.

@torch.cuda.amp.autocast()
./python3.12/site-packages/RealESRGAN/model.py:49: FutureWarning: You are using torch.load with weights_only=False (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for weights_only will be flipped to True. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via torch.serialization.add_safe_globals. We recommend you start setting weights_only=True for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature.

This PR fixes those warnings with two simple changes in model.py:

  • Instead of using the deprecated @torch.cuda.amp.autocast() use: @torch.amp.autocast(device_type='cuda')
    • Update model.py #14 addresses this a different way which might be better for non-CUDA device support
  • The torch.load() call now passes weights_only=True to address code-execution risks during serialization
    • I don't know these models well enough to know if this will break anything, but in my local tests it seems to work fine

Feel free to close this if it doesn't meet the projects standards or is already being addressed elsewhere. I'm sharing in case it is useful to others.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

1 participant