[Feature] TurboMind Engine: Add Vision Encoder support for Qwen3.5 model

[hfassold]

Motivation

Currently it seems that TurboMind Engine does not support the Vision Encoder of Qwen3.5 model.
Is stated in note [3] under the table at
https://lmdeploy.readthedocs.io/en/latest/supported_models/supported_models.html#turbomind-on-cuda-platform

Would be great if support for the Qwen3.5 vision encoder could be added for 'TurboMind' engine.

Related resources

No response

Additional context

No response

[windreamer]

This should already been fixed in #4430 . Could you please try it and feedback to us?

[lvhan028]

Hi, @hfassold
Unfortunately, we don't have plans to implement this in the short term in the same way we did for the Qwen2/2.5 vision encoder. However, we do have a medium-term plan to implement it directly in "src/turbomind", likely targeting late April or May.

[hfassold]

Thanks for the information, looking forward !
@windreamer - see the answer of @lvhan028

[github-actions]

This issue is marked as stale because it has been marked as invalid or awaiting response for 7 days without any further response. It will be closed in 5 days if the stale label is not removed or if there is no further response.

[github-actions]

This issue is closed because it has been stale for 5 days. Please open a new issue if you have similar issues or you have any new updates now.