Improve LayerNorm reciprocal-square-root accuracy - #1667
Open
ollycassidy13 wants to merge 1 commit into
Open
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Improve LayerNorm reciprocal-square-root accuracy
Why
LayerNorm can operate on low-variance inputs where error in the existing reciprocal-square-root approximation becomes large enough to noticeably affect output accuracy in SigLip.
How
This adds an optional second Newton–Raphson refinement:
r = y × (1.5 - 0.5 × x × y²)The refinement is implemented as a pipelined three-FMA FP32 datapath. The LayerNorm configuration exposes
numRsqrtRefinements, defaulting to the existing single-refinement behavior, and updates latency and DSP estimates accordingly. An RTL simulation test verifies the improved result against the ONNX LayerNorm reference.