Top k softmax #175
Open
Top k softmax #175
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
Co-authored-by: Joe Todd <[email protected]>
aacostadiaz
reviewed
Jan 13, 2025
| cutlass_example_add_executable( | ||
| 61_hopper_gemm_with_topk_and_softmax | ||
| 61_hopper_gemm_with_topk_and_softmax.cu | ||
| ) | ||
| else() |
There was a problem hiding this comment.
elseif (SYCL_NVIDIA_TARGET)
There was a problem hiding this comment.
this else if for intel, not nvidia gpu (actually it should be generic).
aacostadiaz
reviewed
Jan 13, 2025
| @@ -193,6 +193,7 @@ struct Sm90TmaWarpSpecializedBiasElementwise { | |||
| #if defined (SYCL_INTEL_TARGET) | |||
| struct IntelPVCEpilogue { | |||
| static constexpr int SubgroupSize = 16; | |||
| static constexpr int FragmentSize = 8; | |||
There was a problem hiding this comment.
What is FragmentSize? Is it always 8 or does it depend on the MMATiled definition?
There was a problem hiding this comment.
It is number of elements processed by one call to epilogue. In xe_epilogue it is defined to depend on MMA atom size.
aacostadiaz
reviewed
Jan 13, 2025
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Outdated suggestions cannot be applied.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.
Implements Top K Softmax for PVC. This includes extending xe_epilogue to call EVT interfaces that are required for this operation and fixing a bug in generic version of Top K Softmax epilogue.