F L U X.1 Fill dev Lo R A extract

Introduction

FLUX.1-Fill-dev-LoRA-extract is a project by SporkySporkness, focused on creating LoRA extracts from the FLUX.1-Fill-dev model available on Hugging Face. These extracts are made using a specific branch of the Kohya_ss repository.

Architecture

The project involves using Low-Rank Adaptation (LoRA) techniques on the FLUX.1-Fill-dev model. It provides various network dimensions to experiment with, though optimal settings such as rank and weight remain undetermined.

Training

LoRA extracts are produced from the FLUX.1-Fill-dev model. Users are encouraged to experiment with different ranks, with a suggestion to use rank 128, similar to the FLUX.1-Depth-dev and FLUX.1-Canny-dev LoRAs.

Guide: Running Locally

1. Clone the repository or download the necessary files from Hugging Face.
2. Install dependencies, including any specific libraries required for running LoRA models.
3. Set up your environment to support GPU acceleration. Using cloud GPUs from providers like AWS or Google Cloud is recommended for optimal performance.
4. Experiment with different ranks and weights to evaluate model outputs.

License

The project is released under the FLUX-1-Dev-Non-Commercial-License. For more details, refer to the LICENSE file in the repository.