[WebNN EP] Fix MatMulNBits when K is not a multiple of block_size - #31152
Merged
Conversation
MatMulNBits stores B as [N, n_blocks_per_col, blob_size] with n_blocks_per_col = ceil(K / block_size), so the dequantized weights span n_blocks_per_col * block_size (padded_K) columns, which is >= K. The previous code reshaped the dequantized operand directly to [N, K], an invalid element-count mismatch whenever K % block_size != 0. Reshape to [N, padded_K] first, then slice off the padding columns to obtain exactly [N, K] before transpose and matmul.
|
Azure Pipelines: There may be pipelines that require an authorized user to comment /azp run to run. |
Contributor
Author
|
@fdwr, PTAL, thanks! |
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Description
MatMulNBitsstoresBas[N, n_blocks_per_col, blob_size]withn_blocks_per_col = ceil(K / block_size), so the dequantized weights spann_blocks_per_col * block_size (padded_K)columns, which is>= K. The previous code reshaped the dequantized operand directly to[N, K], an invalid element-count mismatch wheneverK % block_size != 0.Motivation and Context
Reshape to
[N, padded_K]first, then slice off the padding columns to obtain exactly[N, K]before transpose and matmul.