SDXL Fine Tuning | Worker

 

🚀 | This serverless worker is used for fine tuning SDXL. It supports dynamic auto-scaling using the built-in RunPod autoscaling feature.

Docker Arguments:

1. HUGGING_FACE_HUB_WRITE_TOKEN: Your private Hugging Face token. This token is required for pushing models to the hub.

Build an Image:

docker build -t <your_dockerhub_directory>/image_name:tag --build-arg HUGGING_FACE_HUB_WRITE_TOKEN=<your_token_value> .

Please make sure to replace your_hugging_face_token_here with your actual Hugging Face token to enable model downloads that require it.

Ensure that you have Docker installed and properly set up before running the docker build commands. Once built, you can deploy this serverless worker in your desired environment with confidence that it will automatically scale based on demand.

Test Inputs

The following inputs can be used for testing the model:

{
  "input": {
    "dataset_url": "https://github.com/runpod-workers/sample-inputs/raw/main/images/runpod_dog2.zip",
    "output_directory": "/workspace/output",
    "instance_prompt": "a photo of <subject>",
    "batch_size": 32,
    "training_steps": 1000
  }
}

NOTE: This SDXL fine-tuning worker requires you to provide a folder containing the images of the new instance you'd like to include in the SDXL model. Additionally, we need to ensure that you provide an instance prompt with the following syntax: 'a photo of subject' where "subject" is the name of the item you want to use for fine-tuning the SDXL model (e.g., the images inside the dataset directory).