Lightning-AI / Lightning-AI/lightning-thunder

Sunset `thunder/benchmarks/distributed.py` and Improve `thunder/benchmarks/benchmark_litgpt.py`

Open
#158 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

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

Description

  • [cosmetic] improve the format of JSON output of benchmark_litgpt.py

https://github.com/Lightning-AI/lightning-thunder/blob/cdd43a7fc1110eec10f1854250299b84d1c3b2a8/thunder/benchmarks/distributed.py has been useful but I would find it not easy to extend, e.g. to support gradient accumulation.

https://github.com/Lightning-AI/lightning-thunder/blob/cdd43a7fc1110eec10f1854250299b84d1c3b2a8/thunder/benchmarks/benchmark_litgpt.py would be easy to work with as in https://github.com/Lightning-AI/lightning-thunder/pull/45 which is adding gradient accumulation with no_sync.

cc @crcrpar @carmocca @awaelchli

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 by reading thunder/benchmarks/distributed.py and thunder/benchmarks/benchmark_litgpt.py, then compare the approach with pull request 45 and its gradient-accumulation reference. Done means the distributed benchmark is sunset and benchmark_litgpt.py produces improved JSON output, with the intended format confirmed before implementation.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
performance
Issue type
Refactor
Difficulty
3/5
Estimated time
1-2 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.