-
Notifications
You must be signed in to change notification settings - Fork 89
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
bug(datasets): remove unsafe PyTorch load and save #850
Comments
@gitgud5000 I saw you have another PR with a proposed improvement to the |
Thanks for flagging this, @deepyaman. Perhaps we can skip the check for now (especially since this is still an experimental dataset) and take some time to find a more permanent solution. There's also a possibility that PyTorch may update its methods in the future? |
This is not an experimental dataset, but perhaps we should demote it and ignore the issue until it's resolved properly. |
@merelcht The |
Oh my bad, I thought it was the Yeah let's ping the author. They might be able to give insight into whether this is a real issue or safe to ignore. |
I believe it was @bpmeek! Copying the traceback here for reference -
|
@deepyaman I agree—this is just wrapping the interface for saving/loading in-house objects, not external or intended for distribution... I haven't tried the safetensors with TensorFlow yet, but with Keras > 3, their serialization forces you to use the built-in methods/API anyway. |
I can try to look into this more tomorrow. I'm a little confused though because when I wrote this I was attempting to follow PyTorch's recommended load and save method but perhaps something has changed. |
CI is failing due to https://bandit.readthedocs.io/en/latest/plugins/b704_pytorch_load_save.html, a Bandit rule that was introduced earlier today in their 1.7.10 release.
safetensors
is one possible option for addressing this, but I think somebody who is using TensorFlow with Kedro should ideally weigh in (I don't know what the additional considerations there could be).I'm also not sure if this is a real issue TBH; we're not loading arbitrary data in an application, but rather wrapping an interface that the user could use to load their data.
The text was updated successfully, but these errors were encountered: