benjimon
benjimon
They just released this awesome models weights. It would be great to be able to train it. From what I can tell it uses many aspects of the SDXL pipeline...
It is SDXL architecture with a different noise scheduler(Kolors Scheduler?) and text encoder (ChatGLM3-base). Would be great to be able to fine tune it as it is better than any...
Did you try running either one? Would it be possible to generate a requirements.txt from the working environment with "pip freeze > requirements.txt". Also include your cuda version.