Add AMD Rowwise FP8 Matmul #2611
Closed
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
|
This pull request was exported from Phabricator. Differential Revision: D57600068 |
✅ Deploy Preview for pytorch-fbgemm-docs ready!
To edit notification comments on pull requests, go to your Netlify site configuration. |
Summary: This diff extends the `fp8fp8bf16_rowwise` gemm operation to AMD through a new CK kernel. The new kernel requires new stride support that is only available in developer branches of CK, so we must rely on the `ai_codesign/gen_ai` CK repo. I also extend the fp8 benchmarking suite to include rowwise measurements. I'll soon add detailed benchmarking results but the quick summary is that performance looks quite good, typically inline with tensorwise quantization and sometimes faster, presumably due to using the latest and greatest CK pipelines. Reviewed By: jianyuh Differential Revision: D57600068
|
This pull request was exported from Phabricator. Differential Revision: D57600068 |
|
This pull request has been merged in 8e335d3. |
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Outdated suggestions cannot be applied.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 extends the
fp8fp8bf16_rowwisegemm operation to AMD through a new CK kernel. The new kernel requires new stride support implemented by @zjing14 that is only available in developer branches of CK, so we must rely on theai_codesign/gen_aiCK repo. I also extend the fp8 benchmarking suite to include rowwise measurements. I'll soon add detailed benchmarking results but the quick summary is that performance looks quite good, typically inline with tensorwise quantization and sometimes faster, presumably due to using the latest and greatest CK pipelines.Differential Revision: D57600068