Minimal Model Configuration for LightX2V WAN 2.2 / RTX 4090 (Official Docker / FP8)

[mikezupper]

I'm a developer utilizing the official LightX2V Docker image (lightx2v/lightx2v:25101501-cu128) on a system with an NVIDIA RTX 4090 (host environment is stable and working).

I am seeking the minimal, non-ComfyUI setup to generate videos with WAN 2.2. Specifically, I need precise guidance on the model configuration required for this powerful hardware.

My main confusion is how to structure my local ./models directory and which specific files are necessary to activate FP8 inference for WAN 2.2.

Given the volume mount:
... -v ./models:/workspace/models ...

I need clarification on these specific technical points:

1. Model Configuration for FP8: What are the exact model files (e.g., checkpoints, config files, weight formats, etc.) needed for WAN 2.2, and which specific file ensures the proper configuration to leverage FP8 inference on the RTX 4090?
2. Local Directory Structure: What is the precise, minimal directory structure LightX2V expects for the WAN 2.2 models inside my local ./models folder?

Any streamlined guidance on these two points would be greatly appreciated. Thank you!