MLX LM has moved!

[awni]

MLX LM has moved to a new repo. We will be removing it soon from the examples here as it is now a standalone repo and python package.

• For new issues and discussions related to MLX LM please open them in the new repo.
• For new features / PRs, fork and develop on the new repo as soon as possible to avoid diverging.
• For existing PRs that you would like to move, the simplest approach is likely to fork the new repo and manually apply the changes (cp -r old_repo/llms/mlx_lm new_repo/mlx_lm) should work reasonably well. For more complex diffs for which this doesn't work let us know here or in the PR and we will work with you on moving it.

[Hozefa-L]

Hello,

I want to experiment with pre-trained model like bert. Just to clarify, i can still use mlx-examples for that and only the llms are going to be separated from this repo?
One other question regarding PLMs is i only see the basic encoder and decoder model compatibility in the mlx-examples, what if i want to experiment with other efficient and effective bert-like model such as graphcodebert, should i directly work with the transformers library, since the conversion to mlx might not be straightforward and similar to bert?

Thank you.

[awni]

Just to clarify, i can still use mlx-examples for that and only the llms are going to be separated from this repo?

Yes!

One other question regarding PLMs is i only see the basic encoder and decoder model compatibility in the mlx-examples, what if i want to experiment with other efficient and effective bert-like model such as graphcodebert, should i directly work with the transformers library, since the conversion to mlx might not be straightforward and similar to bert?

I would take it on a case-by-case basis. For some models they may be supported in MLX here or in a third party library. Others may not.