FluxGym Docker Image

Overview

This Docker image provides a flexible environment for running FluxGym, a machine learning project based on the FluxGym and sd-scripts repositories. It offers customizable installation options for different components, allowing users to tailor the image to their specific needs.

Features

• Based on Python 3.10
• Includes FluxGym and sd-scripts repositories
• Customizable installation options for:
  • FluxGym requirements
  • PyTorch (with CUDA 12.1 support)
  • Pre-trained models

Usage

To use this image, you can pull it from Docker Hub:

docker pull thelocallab/fluxgym-flux-lora-training:latest

Examples

1. Run with default settings: docker run -it [your-dockerhub-username]/fluxgym:latest

Components

• FluxGym: [***]
• sd-scripts: [***] (sd3 branch)

Pre-trained Models

When DOWNLOAD_MODELS is set to true, the following models are downloaded:

• UNet: flux1-dev.sft
• CLIP: clip_l.safetensors, t5xxl_fp16.safetensors
• VAE: ae.sft

Note

Please refer to the FluxGym and sd-scripts documentation for more information on how to use these tools.

Support

For issues related to this Docker image, you can reach out to me through my ***.

*** Link - [] *** Link - [] *** Link - [***]

For questions about FluxGym or sd-scripts, please refer to their respective GitHub repositories.

Cocktailpeanut - [] Kohya SS - []