could u give some instructions to support other models? #44
Description
Hi, thanks for great work! This repo really attract a lot of attention from community in recent days!
We are developing new flagship model which size is similar to Z-image-Turbo but with different model structure and much better result.
We have tested RL algorithm like Flow-GRPO and DiffusionNFT based on their original repo code base.
As Flow-Factory provided a unified visual generation post training pipe, we want to try it with our new t2i model and relevant reward models.
I found codebase of Flow-Factory have many modular design. How should I start? Could u add a new part like How to Support Other Models? We will cite this work if we find it is really convenient to fast model version iteration。