bad performance of quant op when enabled EP
Open
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 565
- Forks
- 585
- Avg merge
- 3d 4h
- Merged PRs (30d)
- 366
Description
when enabled EP, each EP rank only compute the tokens routed to itself, but for quant op, it compute all the token(global token)
without deepep, sglang leverage allreduce to perform allgather, each ep rank will get all the tokens
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
No file or test is named. Start by tracing the quant op's token inputs with EP enabled and compare the allreduce/allgather behavior described with and without deepep; done means each EP rank processes only its routed tokens instead of the global token set.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- distributed-systems, performance
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 25/100