Until recently I have never used a LoRA.
So I decided to catch up and get up to speed. You can see the whole journey with details that you can follow here. https://followfoxai.substack.com/p/getting-started-with-loras-vodka
As usual, our post includes every step, tools used, datasets, code, and trained models with the full free access. With our aspiration to create the most open-source-minded series.
A very brief summary:
Start by downloading a few LoRAs and applying them to different models
Then try to train your first model as soon as possible. Following someone else's approach step-by-step is optimal.
For training a LoRA you need three main parts:
Preparing a training dataset
Having a tool or code that will be used to train the model
Choosing the parameters to accomplish your goal.
For your custom LoRA start by setting the goals, then start with copying the methods from other peoples and from there gradually experiment