[QNN-EP] Add missing scale and bias patter to Layernorm fusion with scale =1, bias = 0. #26830
+184
−63
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
Signed-off-by: Mu-Chein Hsu <[email protected]>
quic-muchhsu
changed the title
Signed-off-by: Mu-Chein Hsu <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.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
Layernorm pattern without Mul(scale) and Add(bias) will be fuse to Layernorm(scale =1, bias=nullptr).
If pattern matching breaks at missing Mul(scale), it won't fallback, but instead return Div as last node of the pattern, create a initializer of ones as scale, and return a success Layernorm pattern with scale=1 and bias=nullptr.
Motivation and Context
Layernorm pattern without fusion is very computational expensive. We should consider a missing Scale pattern as Layernorm to have better performance.