How many training images and for how many epochs for training?

#1
by agc2 - opened

Great work! I'm trying to train a shuffle model, but I have to use thousands of images like regular controlnets? Do captions have to contain a trigger word?

Thank you!

I used 921 images for canny, scribble and depth v1 models (it will be better to use more images). I synthesized the dataset. The documentation is here: https://github.com/kohya-ss/sd-scripts/blob/sdxl/docs/train_lllite_README.md

I think ControlNet-LLLite model doesn't require the trigger word.

Sign up or log in to comment