Hello friends! Today we are tackling the lovely world of SimpleTuner. This is an amazing piece of software, always getting support for the latest models first! Sadly it comes at the expense of a GUI and windows support but we are using RunPod for this so it should be smooth sailing.
I've prepared a quick video outlining the process. One issue that came up was that the documentation was a tiny bit outdated based on a commit to the main diffusers branch happening while I was recording, but I figured it out and continued.
My LoRAs come out bad but I think it's a skill diff issue. Hopefully some of you guys can get it right and share your insights.
I'm working on an updated video where I use a better optimizer and settings. Expect it in the coming days. SimpleTuner has undergone a number of updates that basically make LoRA training foolproof.
If you can, consider using my RunPod link so I can keep training
Here is the updated video:
[OLD] Anyway, here is the video: [link]
Here are some relevant links:
FLUX Training Quickstart:
https://github.com/bghira/SimpleTuner/blob/main/documentation/quickstart/FLUX.md
Caption Helper:
https://sd-caption-helper.vercel.app/
https://github.com/markuryy/caption-helper
Latest Hardware Requirements:
https://github.com/bghira/SimpleTuner?tab=readme-ov-file#flux1-dev-schnell
LoRA Testing Space:
https://huggingface.co/spaces/markury/FLUX.1-dev-LoRA
ComfyUI Workflow with LoRA:
https://civitai.com/models/618997