Skip to content

[serge] Fix 10 integration tests for model gemma3 failing with output_mismatch (list output differs (8), other (2)) - #47310

Closed
sergereview[bot] wants to merge 1 commit into
mainfrom
serge/fix/itf-1c12a6ad471e-ccb46e56
Closed

[serge] Fix 10 integration tests for model gemma3 failing with output_mismatch (list output differs (8), other (2))#47310
sergereview[bot] wants to merge 1 commit into
mainfrom
serge/fix/itf-1c12a6ad471e-ccb46e56

Conversation

@sergereview

@sergereview sergereview Bot commented Jul 14, 2026

Copy link
Copy Markdown
Contributor

CI

Original CI failure

  • Failure group: 1/1: 10 integration tests for model gemma3failing withoutput_mismatch (list output differs (8), other (2))
  • tests/models/gemma3/test_modeling_gemma3.py::Gemma3IntegrationTest::test_dynamic_sliding_window_is_default [multi-gpu] (output_mismatch, seen 6/7)

Relates to #47309

Fixes the test_dynamic_sliding_window_is_default failures in the Gemma3 integration group. The generation code from #40039 unsets cache_implementation="hybrid" so that models with sliding-window attention use the dynamic sliding-window cache by default. It was unsetting this value unconditionally, so an explicit cache_implementation="hybrid" passed to generate() was ignored and a DynamicCache containing DynamicSlidingWindowLayer layers was returned instead of the static hybrid cache.

The patch only unsets "hybrid" when it was not explicitly provided in generate() kwargs, preserving user intent.

Note: The normalizer failures (modeling_rules_doc, docstrings) are caused by a missing transformers.models.tipsv2 module / Tipsv2Config import, which is unrelated to this change.


This change was produced automatically by serge from a CI failure report. The patch was generated by an LLM and applied by serge; review before merging.

serge v0.1.0 · model: moonshotai/Kimi-K2.7-Code · 62 LLM turns · 62 tool calls · 194.5s · 2151228 in / 20968 out tokens

…put_mismatch` (list output differs (8), other (2))
@sergereview
sergereview Bot marked this pull request as ready for review July 14, 2026 01:00
@HuggingFaceDocBuilderDev

Copy link
Copy Markdown

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

@github-actions

Copy link
Copy Markdown
Contributor

CI recap

Dashboard: View test results in Grafana
Latest run: 29297459387:1
Result: success | Jobs: 15 | Tests: 171,053 | Failures: 0 | Duration: 14h 20m

@tarekziade

Copy link
Copy Markdown
Collaborator

run-slow: gemma3

@tarekziade
tarekziade requested a review from vasqu July 15, 2026 11:27
@github-actions

Copy link
Copy Markdown
Contributor

Workflow Run ⚙️

This comment contains run-slow, running the specified jobs:

models: ["models/gemma3"]
quantizations: []

@github-actions

Copy link
Copy Markdown
Contributor

CI Results

Workflow Run ⚙️

Commit Info

Context Commit Description
RUN 2ab07868 workflow commit (merge commit)
PR c02bfb6c branch commit (from PR)
main e194978a base commit (on main)

⚠️ Model CI failed to report results

The test failure analysis could not be completed. Please check the workflow run for details.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants