deepspeedai / deepspeedai/DeepSpeed

Is there more detailed documentation for HF AutoTP training?

Open
#7,238 8 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

bug deepspeed-chat
Dominant language
Python
Stars
43.1k
Forks
5k
Avg merge
4d 15h
Merged PRs (30d)
112

Description

For example, the relationship between global_rank and tp_rank / dp_rank, and how to manually all-gather parameters to a specific rank (can we use deepspeed.zero.GatheredParameters ?).
And for example, the list of supported models and how to use this feature with DeepSpeed's native API.
We are planning to try HF AutoTP in OpenRLHF.

Thanks

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 by reading the existing HF AutoTP documentation and DeepSpeed native API guidance, then identify what is missing. Done should clearly explain global_rank, tp_rank, and dp_rank; parameter gathering to a chosen rank; supported models; and use through DeepSpeed's native API.

Written by the indexing model from the issue text.

Assessment

Tech stack
huggingface, python
Domain
distributed-systems, documentation, machine-learning
Issue type
Documentation
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.