meta-pytorch / meta-pytorch/data

[Stateful DataLoader] CI takes too long for Mac and WIndows

Open
#1,338 1 comment 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
1.3k
Forks
179
Avg merge
6d 1h
Merged PRs (30d)
2

Description

🐛 Describe the bug

CI for Ubuntu / Linux in stateful dataloader takes about 20 seconds, but for mac and windows takes about 30+ minutes. Not only burning a huge amount of CI compute budget but slowing down iteration.

Linux probably runs so quickly because it's able to fork a process instead of spawning. The reason not to drop them is if the CI would have caught bugs that would not appear on Linux tests, and since we're not building native code in Stateful DataLoader, I'd assume it's unlikely.

Versions

Nightly

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 locating the CI jobs that exercise Stateful DataLoader on Linux, macOS, and Windows, then compare their runtimes and process behavior. Done means identifying why the non-Linux jobs take 30+ minutes and reducing that cost without losing necessary platform coverage.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
ci-cd, data, performance
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Quiet
Clarity
Mostly clear
Newbie friendliness
45/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.