Add FP8 x INT4 Gemm to Quantize Benchmarks #2984
Open
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
Summary: This diff adds FP8 x INT4 matmul to our quantize benchmarks. I also add a new GB/s value that is collected for all kernels and do some performance tuning of the cutlass F8I4 kernel. Unfortunately, we see in [this benchmarking sheet](https://docs.google.com/spreadsheets/d/1I3h9_JXV7d2spEycvHMv_zACNYEmgDrPzXDJNtpxNPc/edit?usp=sharing) that F8I4 isnt better than F8F8 for any shape, even this set of memory bound shapes that it should favor. We will likely need to follow up with the cutlass team to see if this poor performance is expected. Differential Revision: D60933219
✅ Deploy Preview for pytorch-fbgemm-docs ready!
To edit notification comments on pull requests, go to your Netlify site configuration. |
|
This pull request was exported from Phabricator. Differential Revision: D60933219 |
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Outdated suggestions cannot be applied.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.
Summary:
This diff adds FP8 x INT4 matmul to our quantize benchmarks. I also add a new GB/s value that is collected for all kernels and do some performance tuning of the cutlass F8I4 kernel.
Unfortunately, we see in this benchmarking sheet that F8I4 isnt better than F8F8 for any shape, even this set of memory bound shapes that it should favor. We will likely need to follow up with the cutlass team to see if this poor performance is expected.
Differential Revision: D60933219