This is a reposted model.
I am not the original creator.
Warning: This model is generally only used as a base model for training LoRA, and the image generation effect of this model is not as good as f1.d-fp8.
But if you insist on using this model to generate image, it also works. You can have a try.
source: Kijai/flux-dev2pro-fp8 · Hugging Face
And maybe above is also not the original creator, below is a fp16 version.
source: ashen0209/Flux-Dev2Pro · Hugging Face
I don't know how they are related, so I posted the sources of both.
The principle and effect of this base model are as follows: (also reposted)
source: Why Flux LoRA So Hard to Train and How to Overcome It? | by John Shi | Aug, 2024 | Medium
