SDXL1.0 LoRa Training made easy - in Colab

UPDATE: https://civitai.com/articles/4121/sdxl-lora-training-guide-2024-feb-colab
New article for 2024 with colab link and video walkthrough :)

If you want to train SDXL Lora feel free to use my Fork of Linaqrufs trainer:
https://github.com/MushroomFleet/unsorted-projects/blob/main/Johnsons_fork_230727_SDXL_1_0_kohya_LoRA_trainer_XL.ipynb

You gotta put in your huggingface token as usual, then define the path to your dataset on your google drive. After that remember to set the filename for your Lora.

The Notebook is currently setup for A100 using Batch 30. Using V100 you should be able to run batch 12. Using T4 you might reduce to 8. Keep in mind you will need more than 12gb of system ram, so select "high system ram option" if you do not use A100.

The defaults you see i have used to train a bunch of Lora, feel free to experiment.
I put this up for the early birds! so when it updates, you must go to the Authors site which is linked in the Notebook.