Training data here : https://huggingface.co/datasets/codeShare/lora-training-data/tree/main

LoRa trained with Sigma memes. 10 repeats x 30 epochs x 25 steps = 7500 steps

Try writing some text in your output images. It helps guide the output a lot!

//---//

I could probably write a bit here on how to use the LoRa , but I think the easiest option is to just look at the training data in the link above and adapt your prompting style accordingly.

//--//

Use JoyCaption : https://huggingface.co/spaces/fancyfeast/joy-caption-alpha-two

Saves a lot of time writing a long prompt. If you run out of Huggingface GPU resources feel free to run the joycaption notebook on Colab: https://colab.research.google.com/github/camenduru/joy-caption-jupyter/blob/main/joy_caption_jupyter.ipynb