Repository navigation
ggml-openvino: fix gemma-3 on GPU with stateful execution - #340
Merged
Merged
Conversation
…randRanks The pass unsqueezes the lower-rank operand of an Add/Multiply/Subtract whose operand ranks differ. In gemma-3 the lower-rank operand of the post-attention residual add is the norm output, and unsqueezing it makes the GPU plugin compute the layer wrongly: gemma-3 returns empty answers on GPU with stateful execution. Skip the rewrite when the lower-rank operand is an RMS norm output.
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
#337 added AlignEltwiseOperandRanks, which unsqueezes the lower-rank operand of an Add/Multiply/Subtract whose operand ranks differ. In gemma-3 that operand is the RMS norm output in the post-attention residual add, and the GPU plugin then computes the layer wrongly: gemma-3 returns empty answers on GPU with stateful execution.
The pass now skips the rewrite when the lower-rank operand is an RMS norm output. Qwen3.5 and gemma-4, which the pass fixes, are not affected.
Also updates the validated models table: Qwen3.5 and gemma-4-E2B now pass on GPU with stateful execution.