Recommended approach for finetuning / lora
First of all I want to say thanks for making a model without stealing content. My use is purely personal but I feel better using a model that is not filled with questionable content.
How do I actually fine tune it though or make a lora? For previous 4BAdapt model you had a whole setup available, but not sure where to start with this one.
Do you have some suggestion on which path to take? Will the usual train_dreambooth.py work?
I want to try making both a small lora out of ~50 images and a more involved style fine tuning on a larger dataset.
Thanks a lot for any suggestions
Hi, thanks so much for your thoughtful message and for appreciating the effort we’ve put into building this model responsibly.
It really means a lot to hear that you value using responsible sourced content.
you can use the example from our cookbook https://github.com/Bria-AI/cookbooks/blob/dev/examples/Lora_training_BRIA-3.2.ipynb