Xinyuan Tong comments

Results 27 comments of


                                            Xinyuan Tong

Support precomputed multimodal features for qwen-vl models.

There are still some bugs: And don't forget to Format your code according to the [Code Formatting with Pre-Commit](https://docs.sglang.ai/references/contribution_guide.html#code-formatting-with-pre-commit).

Support precomputed multimodal features for qwen-vl models.

cc @mickqian Could you help review this PR? Thanks!

Frontend language separate reasoning support

/gemini review

model(vlm): mistral 3.1

We should later delete `docs/supported_models/vision_language_models.md` and move mistral part to `docs/supported_models/multimodal_language_models.md`, but it's OK to merge now.

model(vlm): pixtral

MMMU bench result: ``` {'Accounting': {'acc': 0.4, 'num': 30}, 'Agriculture': {'acc': 0.562, 'num': 16}, 'Architecture_and_Engineering': {'acc': 0.333, 'num': 30}, 'Art': {'acc': 0.667, 'num': 30}, 'Art_Theory': {'acc': 0.7, 'num': 30}, 'Basic_Medical_Science':...

关于卤菜的制作方案，如绝味、周黑鸭等制作流程方案

#1504

refactor: minor refactors regarding multimodal processing

Still have some bugs: and

Fix condition for streaming output_ids in tokenizer manager

Now whether the `--stream-output` is set or not, the responses API streaming functions properly.

feat(cli): unify sllm-cli and sllm-serve into one CLI entrypoint

So we decide to remove other commands like replay, fine-tune and encode?

[Reasoning + Structured Output] make reasoning compatible with structured output

/tag-and-rerun-ci