perf: faster fp8->fp16 dequantization for pre sm_90 arch #439
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
yzh119
added a commit
that referenced
this pull request
Aug 13, 2024
🤖 I have created a release *beep* *boop* --- ## [0.1.5](v0.1.4...v0.1.5) (2024-08-13) ### Bugfix * Fix PagedPrefill python api and some typos ([#441](#441)) ([3fff008](3fff008)) * fix prefill kernels' lse result for empty kv-cache ([#440](#440)) ([6ac28f4](6ac28f4)) ### Features * decouple float and int workspace buffer ([#442](#442)) ([a7ee566](a7ee566)) ### Performance Improvements * faster fp8->fp16 dequantization for pre sm_90 arch ([#439](#439)) ([c93f647](c93f647)) ### Acknowledgement We thank contributions and feedbacks from the community: [@comaniac](https://github.com/comaniac), [@hnyls2002](https://github.com/hnyls2002), [@jianfei-wangg](https://github.com/jianfei-wangg), [@Yard1](https://github.com/Yard1). --- This PR was generated with [Release Please](https://github.com/googleapis/release-please). See [documentation](https://github.com/googleapis/release-please#release-please). --------- Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> Co-authored-by: Zihao Ye <[email protected]>
yzh119
added a commit
that referenced
this pull request
Aug 13, 2024
Followup of #439 , use `constexpr` in if conditions so that `BIAS_OFFSET` won't exceed 32 at compile time.
zhyncs
pushed a commit
that referenced
this pull request
Aug 14, 2024
hardware fp8->fp16 fast conversion instruction is not available for sm_80 & sm_89, which makes #420 slow for these architectures. this pr uses marlin's fast fp8->fp16x4 conversion algorithm (copied from vllm project) to accelerate such cases. Co-authored-by: Antoni Baum <[email protected]> Co-authored-by: Cody Yu <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
hardware fp8->fp16 fast conversion instruction is not available for sm_80 & sm_89, which makes #420 slow for these architectures.
this pr uses marlin's fast fp8->fp16x4 conversion algorithm (copied from vllm project) to accelerate such cases.
Co-authored-by: Antoni Baum [email protected]
Co-authored-by: Cody Yu [email protected]