[Bugfix] Stop forcing deprecated use_fast for Qwen3.5

[lesj0610]

Purpose

Transformers v5 deprecates the use_fast image-processor selector, but Qwen3.5 dense and MoE processing currently inherit a Qwen3-VL initialization path that injects use_fast=True even when the caller did not specify it. This produces a startup warning for Qwen3.5 and Qwen3.6 checkpoints.

The explicit default was needed when Qwen3-VL support targeted Transformers 4.x. vLLM now requires Transformers 5.5.3 or newer, where omitting the selector chooses torchvision for Qwen's image processor when available and falls back to pil otherwise. This is the same backend behavior as the former use_fast=True default for Qwen.

Changes

• Override get_hf_processor only for Qwen3.5 dense and MoE processing info so vLLM no longer injects a deprecated default.
• Preserve all caller-provided processor kwargs, including explicit use_fast values.
• Keep the shared Qwen3-VL, Qwen2-VL, ColQwen, and other model-family paths unchanged.
• Add regression coverage for both Qwen3.5 processing-info classes.

A direct replacement with top-level backend="torchvision" is intentionally not used. Qwen3VLProcessor forwards top-level initialization kwargs to both image and video sub-processors, and its video processor does not accept that backend selector.

Validation

• 10 passed in tests/models/multimodal/processing/test_qwen3_5.py.
• 5 passed, 369 deselected in tests/models/test_registry.py -k Qwen3_5.
• A local Qwen3.6 checkpoint produced the same image processor class, video processor class, torchvision image backend, image configuration digest, and video configuration digest before and after omitting the forced default; the deprecated warning was absent only on the new default path.
• ruff check, ruff format --check, py_compile, and git diff --check passed.
• Repository pre-commit hooks passed during commit.

The full vLLM test suite was not run.

[coderabbitai]

Important

Review skipped

Draft detected.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: 000e9ca2-45fc-4024-8997-f82ca26e7c80

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

_{Comment @coderabbitai help to get the list of available commands.}