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

update API to use latest TRL #182

Merged
merged 17 commits into from
Jul 30, 2024
Merged

update API to use latest TRL #182

merged 17 commits into from
Jul 30, 2024

Conversation

kashif
Copy link
Contributor

@kashif kashif commented Jul 28, 2024

fixes #180 and updates the API to use updated transformers and TRL versions #178

@HuggingFaceDocBuilderDev

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

@kashif kashif requested review from edbeeching and alvarobartt July 28, 2024 16:19
Copy link
Member

@alvarobartt alvarobartt left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM thanks for the update @kashif!

Left some minor comments, but IMO we should keep some dataclass fields as those override existing defaults in either trl or transformers, or just ignore those and don't override the existing defaults. Additionally, I saw that the hub_model_revision is only available for DPOConfig so I'd either remove that for DPOConfig or add it to SFTConfig and ORPOConfig too.

Thanks again 🤗

README.md Outdated Show resolved Hide resolved
src/alignment/configs.py Outdated Show resolved Hide resolved
src/alignment/configs.py Outdated Show resolved Hide resolved
src/alignment/configs.py Outdated Show resolved Hide resolved
src/alignment/configs.py Outdated Show resolved Hide resolved
src/alignment/configs.py Outdated Show resolved Hide resolved
src/alignment/configs.py Show resolved Hide resolved
src/alignment/configs.py Show resolved Hide resolved
src/alignment/configs.py Outdated Show resolved Hide resolved
src/alignment/configs.py Show resolved Hide resolved
kashif and others added 4 commits July 29, 2024 16:54
Co-authored-by: Alvaro Bartolome <[email protected]>
Co-authored-by: Alvaro Bartolome <[email protected]>
Co-authored-by: Alvaro Bartolome <[email protected]>
Co-authored-by: Alvaro Bartolome <[email protected]>
@kashif
Copy link
Contributor Author

kashif commented Jul 29, 2024

@alvarobartt do you know where the hub_model_revision is being used?

@alvarobartt
Copy link
Member

@alvarobartt do you know where the hub_model_revision is being used?

So AFAIK we used it in the past to e.g. push different versions of a fine-tune under diff branches, but probably not widely used, so I'd let @lewtun or anyone else chime in! But I believe we can maybe just go ahead with the current changes and leave that for a separate PR in case we want to keep that, WDYT? 🤗 Thanks again for the PR!

@kashif kashif merged commit 95dc472 into huggingface:main Jul 30, 2024
2 of 3 checks passed
@kashif kashif deleted the trl-0.9.6 branch July 30, 2024 07:16
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.

Exception: cannot import name 'log' from 'torch.distributed.elastic.agent.server.api'
3 participants