[sergereview]

Original CI failure

• Failure group: 5 integration tests for model qwen3_vl_moefailing withother, raising RuntimeError (other (5))
• tests/models/qwen3_vl_moe/test_modeling_qwen3_vl_moe.py::Qwen3VLMoeIntegrationTest::test_small_model_integration_test [multi-gpu] (other, seen 21/21)

CI traceback — tests/models/qwen3_vl_moe/test_modeling_qwen3_vl_moe.py::Qwen3VLMoeIntegrationTest::test_small_model_integration_test

(line 311)  RuntimeError: We encountered some issues during automatic conversion of the weights. For details look at the `CONVERSION` entries of the above report!

Where to watch it:

• test_small_model_integration_test — Test dashboard

The 5 Qwen3VLMoeIntegrationTest tests OOM inside from_pretrained while converting experts.gate_up_proj (768 MiB transpose temporary).

Qwen/Qwen3-VL-30B-A3B-Instruct is 30B params ≈ 60 GiB in bf16 against 2×22.3 GiB cards, so device_map="auto" fills each GPU to ~100% (21.94 GiB in use, 368 MiB free) and leaves no headroom for the conversion buffer. Each test also reloaded the whole checkpoint.

Test-side fix, following the glm4v_moe/qwen3_omni_moe pattern (#48776): a lazy get_model() classmethod on the MemoryCleanupMixin class loads the checkpoint once with a 70% per-GPU max_memory cap plus a "cpu" entry and an offload_folder, and the tests call it instead of loading their own copy. No library code touched; expected strings unchanged.

Relates to #50143

---

✅ 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/40967690734
• Failed before the fix (base commit): https://github.com/huggingface/transformers/actions/runs/40963929366

Possibly related

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

• #50037 — Remove expired deprecation cycles (PR, closed, updated 2026-10-29)
• #49846 — Fix failing integration tests on Intel XPU (Arc Pro B60) (PR, open, updated 2026-10-27)
• #41095 — Add LLaVA-OneVision-1.5 model and related configurations (PR, open, updated 2026-10-25)
• #49806 — fix failed test cases for unlimited_ocr model (PR, closed, updated 2026-10-22)
• #49424 — Deprecate old input names (PR, closed, updated 2026-10-21)

---

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: deepseek-ai/DeepSeek-V4.1-Flash-1017:together · 43 LLM turns · 48 tool calls · 180.7s · 1692094 in / 77312 out tokens

[github-actions]

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

run-slow: qwen3_vl_moe

[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.