Fusing a mat mul op followed by a scale op on the CPU #5
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 is useful for Bitnet here we have almost all matricx multiplications be followed by scale operations. As a result, we get a ~2% boost in Bitnet PP performance. Implementation is easy when the matrix multiplication is done by iqk_mul_mat. But if iqk_mul_mat is not implemented for the quant type/architecture, we need to add the scaling to llamafile sgemm and to ggml itself, which is way more messy, so I didn't do it yet. Given that Bitnet is just a niche thing for now, I'll just leave it on a branch for now.
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Outdated suggestions cannot be applied.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.
This is useful for Bitnet here we have almost all matrix multiplications be followed by scale operations.
As a result, we get a ~2% boost in Bitnet PP performance.
Implementation is easy when the matrix multiplication is done by
iqk_mul_mat. But ifiqk_mul_matis not implemented for the quant type/architecture, we need to add the scaling to llamafile sgemm and toggmlitself, which is way more messy, so I didn't do it yet.Given that Bitnet is just a niche thing for now, I'll just leave it on a draft PR for now.