deepspeedai / deepspeedai/DeepSpeed

[BUG] pdsh runner doesn't work with tqdm bar

Open
#6,978 3 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

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

Description

Describe the bug
Training tqdm bar won't show in the rank0 console if pdsh is used as the launcher.

To Reproduce
Run any multi-node training on an example script, the tqdm bar from local rank (localhost) won't show up but show as a blank line.

I assume the root cause it pdsh tries to ssh into even localhost, then the bar cannot be logged through a ssh session somehow.

Expected behavior
Be able to see the tqdm bar.

System info (please complete the following information):

  • OS: Ubuntu 22.04
  • x8 GPUs each node
  • Python version 3.12.8
  • 4 Nodes, same network

Launcher context
PDSH

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 reproducing the issue with the PDSH launcher using a multi-node example script and compare the rank0 console with another launcher. Trace how the local-rank tqdm output is handled when PDSH connects to localhost; done means the tqdm bar is visible in the rank0 console during multi-node training.

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
Mostly clear
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.