vllm_omni.model_executor.models.minimax_h3 ¶

MiniMax H3 model components.

Modules:

Name	Description
`checkpoint`	MiniMax H3 checkpoint resolution for the AR text-encoder stage.
`conditioning`	MiniMax H3 validated text and unified encoder-conditioning contracts.
`encoder`
`encoder_processing`	Shared request preparation and condition codec execution for MiniMax H3.
`long_video`	Shared long-video request limits for encoder and diffusion entry points.
`pipeline`	Two-stage MiniMax H3 topology.
`preprocessing`	Shared MiniMax H3 media normalization and Qwen presentation building.
`reference_video`	MiniMax H3 Ref2VA reference-video preparation.
`text_encoder`	MiniMax H3 text encoder on the vLLM Qwen3-VL model runner.