People always ask how I train my models and while I do not have the endurance currently to write a full new guide, this should give enough of a pointer to massively improve your results if you are still struggling.
I included my inference workflows as well because they can make a massive difference, too. You can train a LoRA well, but because your inference workflow is suboptimal, you think it's bad when it actually isn't. Or you think your LoRA is well trained when it only appears so because you use a massively complicated inference workflow, when in reality, a well-trained LoRA should already work well enough with a basic inference workflow.
Everything you need should be in the .zip file attached to this post.
PS: You do not need to use vast.ai. You can adapt this to your local environment as well. This is just how I train.
Cover image is a placeholder.
If this was helpful to you, consider donating to my Patreon or Ko-Fi!