Train LoRa with Runpod
Hey guys, here is a guide for training models with Runpod.
If you sign up with my referral link we both get 5$ free added to account, this is enough to train 1-2 models.
After you make an account, go to pods and select deploy

Then select GPU, I recommend RTX PRO 6000

Select change template

Search for and select AI Toolkit - Ostris

The default settings work fine, but you can tweak them as you see fit. when you are done select Deploy On-Demand

After a few seconds it should be ready, select HTTP Service

Type "password" to log in

Next click datasets and new dataset

Give it a name, and drag in the images you want to use

Then click on new job, here you set your desired model settings. give it a name, add a trigger word and select your dataset. I like to save every 500, steps of 3000 is fine and you can simply add more steps after training is finished if you need more. I also disable sampling, the rest of the settings I keep at default for 90% of my training

Just click the play button and the models will appear under Checkpoints as they finish.

That's it, remember to stop your pod when you are finished to avoid unnecessary spending.
you should remove watermarks if your images have them. A good site to do this here
If you have any questions post them in the discord discussion.