[sergereview]

Original CI failure

• Failure group: 1 integration test for model deepseek_vl_hybridfailing withoutput_mismatch (list output differs (1))
• tests/models/deepseek_vl_hybrid/test_modeling_deepseek_vl_hybrid.py::DeepseekVLHybridIntegrationTest::test_model_text_generation_batched [single-gpu] (output_mismatch, seen 21/21)

CI traceback — tests/models/deepseek_vl_hybrid/test_modeling_deepseek_vl_hybrid.py::DeepseekVLHybridIntegrationTest::test_model_text_generation_batched

(line 331)  AssertionError: Lists differ: ['You[224 chars]nt:\nThe image depicts a fluffy, light brown a[371 chars]he '] != ['You[224 chars]nt:\nA fluffy animal in a fluffyThe image,The [329 chars]he ']

Where to watch it:

• test_model_text_generation_batched — Test dashboard

Relates to #49580

test_model_text_generation_batched failed with an output mismatch. Root cause is a test-input bug, not hardware drift or a library regression: the test calls apply_chat_template(padding=True) without padding_side="left", so it right-pads and generate() starts decoding from a pad position, making element 0 collapse into a repetition loop (A fluffy animal in a fluffyThe image,The fluffy animal in). The sibling deepseek_vl batched test passes padding_side="left"; the hybrid test dropped it when ported.

A prior patch added the kwarg but left EXPECTED_TEXT[0] stale, so it stayed red. This restores padding_side="left" and updates EXPECTED_TEXT[0] to the coherent left-padded value, which is byte-identical to the passing non-batched test_model_text_generation expectation for the same image+prompt (deterministic greedy output, not a hardware variant). Element 1 (bear) already matched.

---

✅ Verified on GPU

serge ran the targeted @slow test(s) on a GPU runner and opened this PR only after they passed with this patch. Each targeted test was run 5× on both the pre-patch and patched trees to rule out flakiness — the result held on every run.

• Passes with this patch: https://github.com/huggingface/transformers/actions/runs/38499493968
• Failed before the fix (base commit): https://github.com/huggingface/transformers/actions/runs/38492780162

Possibly related

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

• #48171 — Fix stale expected values in integration tests (cuda sm_86 / Aug04 regressions) (PR, closed, updated 2026-08-21)
• #44739 — fix series of failed test case for janus model (PR, closed, updated 2026-04-01)
• #44179 — Models with incorrect tokenizer_class in tokenization_config.json tha… (PR, closed, updated 2026-02-23)
• #41977 — Add Phi3.5 Vision Model (PR, open, updated 2025-12-18)

---

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: Qwen/Qwen3.8-27B:novita · 17 LLM turns · 18 tool calls · 1708.4s · 653483 in / 103655 out tokens

[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.

[github-actions]

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

run-slow: deepseek_vl_hybrid