[model] fix: add Qwen3.5 VL to _TEXT_TO_VL_FAMILY for unified checkpoint support - #7433

What does this PR do?

Fixes continuous token builder creation for Qwen3.5 VL unified checkpoints. Their HF model_type is qwen3_5 / qwen3_5_moe (same as the text models), so the wiring resolves them to the text family QWEN35. Since QWEN35 was missing from _TEXT_TO_VL_FAMILY, passing a multimodal processor raised ValueError: Model resolved to the text Continuous Token family 'qwen35', but a multimodal processor was provided, making Qwen3.5 VL unusable for multimodal RL training. This PR adds the QWEN35 -> QWEN3_VL upgrade entry, mirroring the existing DEFAULT -> VL_DEFAULT and GEMMA4 -> GEMMA4_VL entries.

Checklist Before Starting

• Search for similar PRs. Paste at least one query link here: https://github.com/verl-project/verl/pulls?q=is%3Apr+qwen3_5+continuous+token
• Format the PR title as [{modules}] {type}: {description} (This will be checked by the CI)
  • {modules} include fsdp, megatron, veomni, sglang, vllm, rollout, trainer, ci, training_utils, recipe, hardware, deployment, ray, worker, single_controller, misc, perf, model, algo, env, tool, ckpt, doc, data, cfg, reward, fully_async, one_step_off
  • If this PR involves multiple modules, separate them with , like [megatron, fsdp, doc]
  • {type} is in feat, fix, refactor, chore, test
  • If this PR breaks any API (CLI arguments, config, function signature, etc.), add [BREAKING] to the beginning of the title.
  • Example: [BREAKING][fsdp, megatron] feat: dynamic batching

Test

Added a CPU unit test test_qwen35_unified_with_processor_upgrades_to_vl in tests/utils/test_continuous_token_on_cpu.py (class TestWiringVLFactory), parametrized over qwen3_5 and qwen3_5_moe. It asserts that a Qwen3.5 unified checkpoint plus a multimodal processor now resolves to QwenVLContinuousTokenBuilder with supports_multimodal() == True. The test fails on main with the ValueError above and passes with this fix. Also verified end-to-end locally by launching multimodal RL training with a Qwen3.5 MoE unified checkpoint, which previously crashed at builder creation.

API and Usage Example

No API change. Existing entrypoint now works for Qwen3.5 VL unified checkpoints:

Design & Code Changes

• verl/utils/tokenizer/continuous_token_wiring.py: add ContinuousTokenModelFamily.QWEN35: ContinuousTokenModelFamily.QWEN3_VL to _TEXT_TO_VL_FAMILY (one-line change, follows the established upgrade-map pattern).
• tests/utils/test_continuous_token_on_cpu.py: new parametrized unit test covering the upgrade path.

Checklist Before Submitting

• Read the Contribute Guide.
• Apply pre-commit checks: pre-commit install && pre-commit run --all-files --show-diff-on-failure --color=always
• Add / Update the documentation. (Not needed: one-line internal wiring fix, no user-facing behavior documented.)
• Add unit or end-to-end test(s) to the CI workflow to cover all the code.
• Once your PR is ready for CI, send a message in the ci-request channel in the verl Slack workspace.
• If your PR is related to the recipe submodule, please also update the reference to the submodule commit. (Not related.)