Lightning-AI / Lightning-AI/lightning-thunder

`TORCH_NCCL_AVOID_RECORD_STREAMS` is the default

Open
#2,405 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
1.5k
Forks
121
PR merge metrics
No merged PRs in 30d

Description

Now PyTorch avoids nccl record streams by default as per https://github.com/pytorch/pytorch/blob/8ab5868a2199fe485c2d66533b9244ccb97e487d/torch/csrc/distributed/c10d/ProcessGroupNCCL.cpp#L914-L917, the following three may want to be updated.

- https://github.com/Lightning-AI/lightning-thunder/blob/5c18b33917586e388e9526eb28cfd48cf8755915/thunder/distributed/__init__.py#L40
- https://github.com/Lightning-AI/lightning-thunder/blob/5c18b33917586e388e9526eb28cfd48cf8755915/thunder/benchmarks/benchmark_peft.py#L28
- https://github.com/Lightning-AI/lightning-thunder/blob/5c18b33917586e388e9526eb28cfd48cf8755915/thunder/benchmarks/benchmark_litgpt.py#L64

Contributor guide

No contributing guide indexed for this repository

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 the three referenced locations: thunder/distributed/__init__.py, thunder/benchmarks/benchmark_peft.py, and thunder/benchmarks/benchmark_litgpt.py, then compare their NCCL record-stream assumptions with the linked PyTorch behavior. Update the affected references so they match the current default and verify that the distributed module and both benchmarks remain consistent.

Written by the indexing model from the issue text.

Assessment

Tech stack
python, pytorch
Domain
distributed-systems
Issue type
Refactor
Difficulty
2/5
Estimated time
1-3 hours
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
55/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.