THUDM / THUDM/slime

Running train_async.py and removing --colocate makes the train hang

Open
#699 2 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
8.5k
Forks
1.3k
Avg merge
5h 36m
Merged PRs (30d)
22

Description

Dear colleagues,

I am running Slime 0.1.0 for Qwen 30B A3B Thinking on Slurm with 16 H100. When I use train_async.py and remove --colocate, the job hangs after printing the job parameters (word_size, yaml_cfg, etc.) whereas if I run it with sync train, it works just fine.

I have no clue where to look at. Any suggestions are more than welcome. Thank you so much!

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 train_async.py and compare its startup path with the working synchronous training command. Reproduce the hang on Slurm using the stated 16 H100 setup, capture the last job-parameter output, and determine where asynchronous training stops progressing. Done means the non-colocated async command starts or reports a clear actionable failure instead of hanging.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
distributed-systems, machine-learning
Issue type
Bug
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.