Hidream and Hunyuan t2v,i2v, and animatediff also kolors and playground 2.5 and wangp also Hunyuan t2img also please add tea cache and wave speed.

[KintCark]

Will we be getting these model supported soon especially animatediff I've Been waiting for this for a long time 👀👨🏽‍💻😁

[stduhpf]

I can see Hunyuan t2img being supported eventually, and maybe things like TEA cache / wave speed can be implemented without too much hassle.

HiDream and Kolors could also maybe be added at some point, but it's a lot of work to implement support for the autoregressive LLMs as text encoders. The best way to do it would probably be to include llama.cpp to this project for much better language model support?

For the video models it's more tricky, most of them use 5D tensors, and GGML only supports up to 4D tensors for now.