Could the author release a training script for training LoRA or fine-tuning based on Diffusers？
Hello, thank you for your outstanding open-source contribution! We are the InstantX team behind InstantID, and we have taken note of AuraFlow's impressive text-to-image capabilities. We are interested in adapting this model, along with other plugins from InstantX.
Could you possibly release a training script for training LoRA or fine-tuning based on diffusers? With such a script, we could quickly adapt to the model, and it would also be beneficial for developers who are familiar with diffusers to rapidly expand the AuraFlow ecosystem.
it is here: https://github.com/bghira/simpletuner
see auraflow quickstart document
He needs to list that as his information only lists the various Stable Diffusion versions. Not even sure how I would train kolors, or auraflow with it.
Oh, thank you.
https://github.com/bghira/SimpleTuner/blob/main/documentation/QUICKSTART.md
like this?
Thank you for the reply. After taking a look, can we draw the conclusion that the training of AuraFlow and SD3 is completely identical? If they are consistent, we can directly reuse the strategy of SD3 to quickly start the experiment.
nope, they handle text embeds very differently. and the AuraFlow loss calculation is different.
simpletuner has dropped support of auraflow. Does the auraflow team have alternative scripts?