Enable multiple of 16 padding for INT8 Tensor Cores #1982
Merged
Conversation
This file contains hidden or 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
Contributor
Author
|
Confirmed that it works on RTX 5060 Ti: #1981 (comment) |
|
Does the new CUDA Release 13.1 Update 1 fix this problem or is it another issue ? from release notes : Fixed an issue where cublasLtMatmul with FP8 arguments and CUBLASLT_MATMUL_MATRIX_SCALE_SCALAR_32F scaling mode (default) incorrectly required scaling factor addresses to be 16-byte aligned. This issue existed since cuBLAS 12.9. [5728938] |
Contributor
Author
|
@easyfab This is another issue. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
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Outdated suggestions cannot be applied.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.
Enable multiple of 16 padding for INT8 Tensor Cores.
Re-enable INT8 for sm120 - Blackwell GPUs ( revert #1937 )
Thanks @sssshhhhhh for the hint. 👍
Closes: #1981 #1865
INT8 padding was disabled in #332 for reason
it appears padding to a multiple of 16 does not help.I guess, "does not help" meant that padding didn't increase performance.
Tests:
Tested older GPU (Turing) with INT8 Tensor Cores: vocab (
n[fw]) is padded, no change in performance (as I remember, ~4% of GEMM calls needed padding).I don't have Blackwell GPU at hand to test, but the padding should fix "CUBLAS_STATUS_NOT_SUPPORTED" error.