[Feature] Question about Multi-Platform Support

Hi there,

I'm a developer actively using Ascend NPU for running generative AI models, and I've been really impressed by DiffSynth-Studio's design and functionality. Recently, we've successfully adapted several diffusion-based models to run on Ascend hardware, including:

• Wan 2.2
• Wan 2.1
• Stable Diffusion 3.5 Large
• Qwen-Image
• CogVideoX
• etc...

The results show solid performance and stability, which makes me even more excited about bringing open-source Diffusion model engines like DiffSynth-Studio to the Ascend platform.

I'm opening this issue to ask: Are there any plans to support multi-platform execution (e.g., beyond CUDA, to include NPUs like Ascend) in the roadmap?

As a user deeply invested in the Ascend ecosystem, I'd love to help with the integration — whether it's adapting core components, adding backend abstractions, or contributing testing and documentation. I'm eager to collaborate with the community and contribute code to make DiffSynth-Studio run smoothly on more hardware platforms.

Would love to hear your thoughts and guidance on how we could move this forward 🙏

The issue names no repository files, tests, or entry points. Start by reviewing the project's current CUDA execution approach and discussing the intended Ascend backend scope with maintainers. Done would require an agreed implementation plan covering supported platforms, integration boundaries, testing, and documentation.