lucataco / diffusers-dreambooth-lora-x2

FLUX.1-Dev LoRA Training (with 2x GPUs) by Huggingface Diffusers

  • Public
  • 71 runs
  • GitHub
  • Weights
  • License
Run with an API
Playground API Examples README Versions

Run time and cost

This model runs on 2x Nvidia A100 (80GB) GPU hardware. We don't yet have enough runs of this model to provide performance information.

Readme

Huggingface Diffusers Dreambooth LoRA Trainer for Flux.1-Dev

About

Cog implementation of the Diffusers Dreambooth LoRA Trainer

Follow me on Twitter/X @lucataco93

How to use

Upload Images: Put all your images in a zip file. This is what you’ll use to train your LoRA model.

Training Steps: When setting the number of training steps, use half of what you normally would. For example, if you usually train a model with 1000 steps on a single GPU, use 500 steps when training with this model which uses two GPUs.

Upload to Huggingface: If you want to upload your trained LoRA model to Huggingface, make sure you have a Huggingface token with write access. You’ll also need to choose a location for your model. You can see an example of an uploaded LoRA model here: lucataco/flux-queso.

Track Training with wandb: You can also use wandb to track your training progress. This can help you monitor how well your model is learning. You can see an example of some wandb tracked runs here

Run your LoRA

Copy the URL from the trained_model.tar from the prediction output, and use the lucataco/flux-dev-lora Explorer model to run your LoRA

An example of a prediction run with a LoRA can be found here queso

Licensing and commercial use

If you generate images on Replicate with FLUX.1 models and their fine-tunes, then you can use the images commercially.

If you download the weights off Replicate and generate images on your own computer, you can’t use the images commercially.