Model Selection for infering training samples

The model used for inference to create the Lora training samples makes some weird results and it's hard to ascertain the quality of the lora epochs. It would be useful if some settings could be changed in the advanced settings panel like selecting a model from a list of your pre-loaded models and setting up some prompts that will end up being used when the lora samples are created at the end of each epoch.

Great feature, by the way. It made me instantly turn off my colab trainer and become a Civitai supporter as soon as I found out I can train loras on your website!