How to custom train Stable Diffusion?

I’ve been trying to mess around with Colab but the process is just a pain. What’s the best way to do this right now?