Lightning-AI / Lightning-AI/lightning-thunder
[Tensor Parallelism] Improve comm optimization logic for pair of column-wise parallel linear and row-wise parallel linear
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 1.5k
- Forks
- 121
- PR merge metrics
- No merged PRs in 30d
Description
as the todo comment of
https://github.com/Lightning-AI/lightning-thunder/blob/24e81cf85f41ccd9dee2d1641a15b91aec65d34d/thunder/tests/distributed/test_tensor_parallel.py#L297-L300
says, currently https://github.com/Lightning-AI/lightning-thunder/blob/24e81cf85f41ccd9dee2d1641a15b91aec65d34d/thunder/distributed/tensor_parallel/optimize_comm.py#L28 is not capable of identifying and removing pairs of post-processing of column-wise parallel linear and pre-processing of row-wise parallel linear on some occasions
Contributor guide
No contributing guide indexed for this repository
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Read the TODO in thunder/tests/distributed/test_tensor_parallel.py around lines 297-300, then inspect thunder/distributed/tensor_parallel/optimize_comm.py around line 28. Trace the existing communication optimization logic and use the tensor-parallel test to verify that the post-processing and pre-processing pair is identified and removed in the relevant cases.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python, pytorch
- Domain
- distributed-systems, performance
- Issue type
- Feature
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Mostly clear
- Newbie friendliness
- 45/100