Torch-based GEMM with fake dequantization for mxfp6 e2m3 #762
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
- Packs f32 into fp6 and unpacks back to f32 for matmul - Pytests for quantization error, matmul error, and LUT rounding - Does not use CDNA4 FP6, only fake dequantization Signed-off-by: Nirmal Senthilkumar <[email protected]>
Giuseppe5
reviewed
Jan 27, 2026
| # Calculate block-wise scales (Microscaling) | ||
| x_reshaped = x.view(-1, block_size) | ||
| amax = x_reshaped.abs().max(dim=1, keepdim=True).values | ||
| scales = amax / MAX_E2M3 |
There was a problem hiding this comment.
This is supposed to be a power-of-two value
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 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Outdated suggestions cannot be applied.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.
Will soon work on adding CDNA4 native mxfp6 operations