[Feature]: Integrate with XGrammar for zero-overhead structured generation in LLM inference.

[choisioo]

🚀 The feature, motivation and pitch

Enables zero-overhead structured generation in LLM inference.

https://github.com/mlc-ai/xgrammar

Alternatives

No response

Additional context

https://blog.mlc.ai/2024/11/22/achieving-efficient-flexible-portable-structured-generation-with-xgrammar
https://xgrammar.mlc.ai/docs/how_to/engine_integration.html

Before submitting a new issue...

• Make sure you already searched for relevant issues, and asked the chatbot living at the bottom right corner of the documentation page, which can answer lots of frequently asked questions.

[dongxiaolong]

Waiting for the following PR to be merged:
#10785
#10803

[mgoin]

Support has landed as of #10785 and performance is much better!