integration with vLLM-Omni

[hsliuustc0106]

Hi, recently vLLM-Omni has been released to support non-text outputs and non-AR arch. Is it possible to get this fantastic work integrated with vLLM-Omni which finally will speed up the dLLM inference seamlessly aligned with vLLM?

[hsliuustc0106]

https://github.com/vllm-project/vllm-omni

[drewjin]

Thanks for your attention, as far as I concerned, vLLM-Omni does not support dLLM inference yet, maybe you can pay more attention at https://github.com/zhijie-group/Diffulex