mosaicml / mosaicml/streaming

clean_stale_shared_memory duplicating the master process when called in a train.py script

Open
#663 2 comments 2 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

bug
Dominant language
Python
Stars
1.6k
Forks
206
PR merge metrics
No merged PRs in 30d

Description

To reproduce

calling clean_stale_shared_memory() at the beginning of a train.py script itself launched with composer in a distributed setup.

Expected behavior

The memory is cleaned at the beginning of the training and then the training happens normally
image

What I get:
image
The process is duplicated on the GPU:0 and is never destroyed

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 tracing clean_stale_shared_memory() from the train.py entry point when launched with Composer in a distributed setup. Reproduce the reported GPU:0 process duplication, then determine why the extra process is not destroyed and verify that shared memory is cleaned before training proceeds normally.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
distributed-systems
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.