How does the Qwen3.5 model disable thinking?

[SongXiaoMao]

lmdeploy serve api_server /home/cheng/model/Qwen3.5-27B-AWQ
--tp 4
--cache-max-entry-count 0.8
--log-level INFO
--max-concurrent-requests 4
--model-name Qwen3.5-27B-AWQ
--backend turbomind
--max_batch_size 64
--api-key abc123
--server-port 8000
--cache-block-seq-len 32 \

[SongXiaoMao]

This one doesn't seem to recognize the image request, how do I need to enable multimodality?

[lvhan028]

Turbomind engine hasn't support Qwen3.5 vision encoder yet.
You can use pytorch engine instead. But unfortunately the multimodal feature was accidently disabled in v0.12.2.
Please use the latest main branch.

[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.