Fix QK norm reshape to match RMSNorm's expected 3D input shape#1221
Open
brendanlong wants to merge 1 commit intoTransformerLensOrg:devfrom
Open
Fix QK norm reshape to match RMSNorm's expected 3D input shape#1221brendanlong wants to merge 1 commit intoTransformerLensOrg:devfrom
brendanlong wants to merge 1 commit intoTransformerLensOrg:devfrom
Conversation
RMSNorm.forward() has a jaxtyping hint expecting a 3D tensor (batch, pos, length), but _apply_qk_norm was reshaping to 2D (batch*pos*heads, d_head), causing a BeartypeCallHintParamViolation on Gemma 3 models. Reshape to 3D (batch*pos, heads, d_head) instead — normalization is identical since RMSNorm only operates on the last dimension. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
7 tasks
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Description
RMSNorm.forward() has a jaxtyping hint expecting a 3D tensor (batch, pos, length), but _apply_qk_norm was reshaping to 2D (batchposheads, d_head), causing a BeartypeCallHintParamViolation on Gemma 3 models. Reshape to 3D (batch*pos, heads, d_head) instead — normalization is identical since RMSNorm only operates on the last dimension.
This is a breaking change because it changes what gets passed to hooks. I'm unsure if we want to fix this or change the type hint, but this PR fixes the output to match the type hint.
Type of change
Checklist:
Note: I have new tests I'm working on in brendanlong#8 that failed due to this