Skip to content

Fix attention handling for hybrid models - #463

Merged
ArtemVazh merged 4 commits into
mainfrom
fix/qwen3.5
Jul 30, 2026
Merged

Fix attention handling for hybrid models#463
ArtemVazh merged 4 commits into
mainfrom
fix/qwen3.5

Conversation

@ArtemVazh

@ArtemVazh ArtemVazh commented Jul 30, 2026

Copy link
Copy Markdown
Collaborator

Handle missing attention layers and correctly map full-attention outputs. Add Qwen no-thinking configuration.

HYDRA_CONFIG="$PROJECT/examples/configs/polygraph_eval_gsm8k_simple_instruct.yaml" \
  "$PYTHON" -u "$PROJECT/scripts/polygraph_eval" \
    model=qwen-no-thinking \
    model.path=Qwen/Qwen3.6-27B-FP8 \

@ArtemVazh
ArtemVazh requested a review from IINemo July 30, 2026 16:12
@ArtemVazh
ArtemVazh merged commit 98dd675 into main Jul 30, 2026
1 check passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@IINemo IINemo IINemo approved these changes

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

2 participants

@ArtemVazh @IINemo