lora
lora copied to clipboard
Custom styles from config files
trafficstars
Hey 👋
I'd like to propose a basic configuration file that substitutes the current static template array used by lora_pti to apply pivotal training. I'm thinking in something simple like a json file that users can tune for their own training jobs.
Happy to tackle this one if @cloneofsimo you think it makes sense and would be beneficial for the project.
Seems like a good idea. I've been using a longer set of prompt templates from the author of the textual inversion paper. I've never explored the effects of changing these templates, so separating it out from the code could make it easy to do some experiments!