[Feature] Profiling GeMM kernel in lmdeploy
Open
- Dominant language
- Python
- Stars
- 8.1k
- Forks
- 748
- Avg merge
- 6d 2h
- Merged PRs (30d)
- 54
Description
### Motivation
Single kernel cost measurement is crucial to the holistic performance of the LLM. I wonder if lmdeploy supports benchmarking either solely quantized, int8, or int16 gemm operations?
Thanks!
### Related resources
_No response_
### Additional context
_No response_
Contributor guide
Assessment
This issue has not been assessed yet.