LAION-AI / LAION-AI/Open-Assistant

accelerator version issue

Open
#3,326 2 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

bug ml
Dominant language
Python
Stars
37.4k
Forks
3.3k
PR merge metrics
No merged PRs in 30d

Description

huggingface's accelerator updated from v0.19.0 to v0.20.0 and 'logging_dir' disappeared from init method in Accelerator class.

So the above error occurs.

image
https://github.com/huggingface/accelerate/blob/baebae3bbecbea05d721a50917f352cccd14811e/src/accelerate/accelerator.py#L242-L244

OA doesn't specify version
https://github.com/LAION-AI/Open-Assistant/blob/0fcf3e08fe62295d4696e590005b0f33383342ea/model/pyproject.toml#L12-L13

However, if you go to the trlx library that actually runs the accelerator, it is versioned as shown below.
image
https://github.com/CarperAI/trlx/blob/0dce99d96b7d70b6a9114129d8e38bf6c80eb653/requirements.txt#L1-L2

Of course, it is true that the trlx library also has its own errors.

However, if OA will have a dependency on the trlx library, I think it is necessary to get the trlx requirement.txt and install it with the same version as specified there.

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start with model/pyproject.toml and compare its dependency declarations with trlx's requirements.txt, then review the linked Accelerator constructor change. The issue is done when the dependency versions are aligned so the reported logging_dir error no longer occurs during installation or use.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
build-system, machine-learning
Issue type
Bug
Difficulty
2/5
Estimated time
1-3 hours
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.