[sergereview]

Original CI failure

• Failure group: 2 integration tests regressed by commit 83d46aa2a2c4 (PR #47625)
• tests/models/minicpmv4_6/test_modeling_minicpmv4_6.py::MiniCPMV4_6IntegrationTest::test_small_model_vision_generation_batch [multi-gpu] (output_mismatch, seen 7/7)

CI traceback — tests/models/minicpmv4_6/test_modeling_minicpmv4_6.py::MiniCPMV4_6IntegrationTest::test_small_model_vision_generation_batch

(line 505)  AssertionError: Lists differ: ['The animal in the image is a Pystylus, a[145 chars] It'] != ["The animal in the image is a Pystylus, a[161 chars]s a"]

Where to watch it:

• test_small_model_vision_generation_batch — Test dashboard

The integration test test_small_model_vision_generation_batch regressed because the Qwen3.5 text-model changes in PR #47625 altered the generated text for this model. The actual decoded output now matches the previously separate ("cuda", 10) expectation, so the ("cuda", None) expected string is updated to the current model output.

Relates to #48579

Possibly related

Existing issues/PRs mentioning test_small_model_vision_generation_batch (keyword match — not verified to share a root cause):

• #48406 — [MiniCPMV4_6] Update test_small_model_vision_generation_batch expected output (value drift) (PR, closed, updated 2026-09-01)

---

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 · 59 LLM turns · 63 tool calls · 83.2s · 2085622 in / 3934 out tokens

[github-actions]

[For maintainers] Suggested jobs to run (before merge)

run-slow: minicpmv4_6

[github-actions]

---

CI recap

Dashboard: View test results in Grafana
Latest run: 34064564900:1
Result: success | Jobs: 2 | Tests: 21 | Failures: 0 | Duration: 2m 23s

[HuggingFaceDocBuilderDev]

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.

[zucchini-nlp]

This again raises the issue of keeping a dependency list for multimodal models, so that when a backbone gets an update then all VLM/ALM that use it as backbone also trigger tests. And they are added in run-slow hint by the bot

I know YD mentioned it already, just re-raising in case it got lost :)

[tarekziade]

This again raises the issue of keeping a dependency list for multimodal models, so that when a backbone gets an update then all VLM/ALM that use it as backbone also trigger tests. And they are added in run-slow hint by the bot

I know YD mentioned it already, just re-raising in case it got lost :)

cc @ydshieh for tracking

[ydshieh]

well, I don't even remember I mentioned this 😆 But good point

[ydshieh]

The expected output values in this PR is incorrect.

The issue was originally fixed https://github.com/huggingface/transformers/pull/48406/changes, the correct value should be

"The animal in the image is a Pystylus, also known as a Pystylus cat or Eurasian pystylus. It",

@remi-or however introduced ("cuda", 10) and having some problems, so a new PR is opened #48454

but it's not merged before.

I will handle it. But @tarekziade what is the hardware type used by serge?