RFC: Modeling Support for Hunyuanvideo Model

Metadata

Item	Details
Status	Approved
Author	wqh17101
Creation Date	2025-12-19
Related Links	https://gitcode.com/Ascend/msit/pull/4911

---

1. Oversview

This proposal aims to address the adaptation issues of the Hunyuanvideo multimodal model, enabling it to run correctly under the tensor_cast framework.

2. Detailed Design

Model Input Construction Adaptation

To support model structure parsing and graph construction under the tensor_cast framework, placeholder inputs conforming to the forward interface of HunyuanVideo series models need to be provided, using tensors with device="meta".Input generation logic has been implemented for the following two types of models:

• HunyuanVideoTransformer3DModel: Constructs the basic encoder_attention_mask;

3. Implementation Plan

Completed

• Completed DIT modeling for hunyuanvideo

Follow-up Development

• Completed DIT modeling for hunyuanvideo1.5
• Support for image input
• VAE modeling
• Diffusers pipeline loading support
• [ ]Support for various parallelization features