Samit

[RFC][WIP]: Support diffusers and community acceleration backends

3

### Motivation. 1. Integrate `diffusers` DiffusionPipeline to easily support popular diffusion models like QwenImage, Flux, Wan2.2 in `diffusers` 2. Support `cache-dit`/`fastvideo` diffusion accelerator to boost diffusion stage throughput. ### Proposed...

[RFC]: Optimize model yaml config

3

### Motivation. Current model yaml config is related to resource allocation (`devices`, `gpu_memory_utilization`), engine config, and model config (Qwen2_5OmniForConditionalGeneration), for example: https://github.com/vllm-project/vllm-omni/blob/main/vllm_omni/model_executor/stage_configs/qwen2_5_omni.yaml It's quite time-consuming to configure. 1. To fill...

[New Model]: Add HunYuanImage3.0

8

### The model to consider. Support HunYuanImage3.0 from https://huggingface.co/tencent/HunyuanImage-3.0 and https://huggingface.co/tencent/HunyuanImage-3.0-Instruct ### The closest model vllm-omni already supports. QwenOmni. The main difference: - HunYuanImage3.0 supports multi-modal understanding and generation. These...

[Docs] Fix get started and installation docs

2

PLEASE FILL IN THE PR DESCRIPTION HERE ENSURING ALL CHECKLIST ITEMS (AT THE BOTTOM) HAVE BEEN CONSIDERED. ## Purpose ## Test Plan ## Test Result --- Essential Elements of an...