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

Please support numpy 2.0 #31802

Open
MonolithFoundation opened this issue Jul 5, 2024 · 6 comments
Open

Please support numpy 2.0 #31802

MonolithFoundation opened this issue Jul 5, 2024 · 6 comments
Labels
Feature request Request for a new feature

Comments

@MonolithFoundation
Copy link

Feature request

As numpy 2.0 comes out, it brings impressive speed up, please consider add numpy 2.0 support.

Besides, numpy is so foundamental, I think all libs should force to align with it rather than force users with lower numpy version/

Currently support numpy 2.0 libs:

  • torch

Thoese don't:

  • tensorflow
  • transformers

So, please just ignore tensorflow, nowadays we don't use it any longer,

Motivation

At least make us have an option.

Your contribution

don't force to lower version, just keep up to date.

@MonolithFoundation MonolithFoundation added the Feature request Request for a new feature label Jul 5, 2024
@LysandreJik
Copy link
Member

cc @ydshieh

@ydshieh
Copy link
Collaborator

ydshieh commented Jul 5, 2024

Hi.

Our internal check indeed shows that transformers works (quite) well with numpy==2.0.0 in a torch-only environment (with a few exception).

However, at this moment, we would prefer users to install numpy==2.0.0 manually and see how it goes with transformers in their daily usage.

We are, of course, happy to remove the pin of numpy version after a period of time of waiting.

Related
#31568
kyamagu/faiss-wheels#104

@MonolithFoundation
Copy link
Author

MonolithFoundation commented Jul 5, 2024 via email

@ydshieh
Copy link
Collaborator

ydshieh commented Jul 5, 2024

@MonolithFoundation

I don't understand well. Could you describe the situation in a bit more detail? Ideally something that we can reproduce.

@ydshieh
Copy link
Collaborator

ydshieh commented Jul 5, 2024

I think I know the situation now. I will check 🤗

@Fred-Wu
Copy link

Fred-Wu commented Jul 14, 2024

Same here.

"ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
transformers 4.42.4 requires numpy<2.0,>=1.17, but you have numpy 2.0.0 which is incompatible."

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
Feature request Request for a new feature
Projects
None yet
Development

No branches or pull requests

4 participants