flux.1-schnell

FLUX.1-schnell

FLUX.1 suite is a set of text-to-image models that define a new state-of-the-art in image detail, prompt adherence, style diversity and scene complexity for text-to-image synthesis, developed by Black Forest Labs.

FLUX.1 [schnell] is the fastest model among FLUX.1 family, which is a 12 billion parameter rectified flow transformer. Trained using latent adversarial diffusion distillation, FLUX.1 [schnell] can generate high-quality images in only 1 to 4 steps. Also, the code and weight is released under the apache-2.0 licence, so that the individual can generate the high quality with FLUX.1 [schnell] in any purpose.

Running Locally

Note: You have to use a GPU machine with CUDA to run FLUX.1 model.

1. Prepare environment. Using virtual environment is strongly recommended.

   python -m venv .venv
   
   source .venv/bin/activate
   pip install -r requirements.txt

2. Launch the app. If the VRAM is not sufficient, add the --offload flag.

   python app.py --offload

3. Access the Gradio interface by opening http://localhost:7860 in your web browser.

Deploying with VESSL Run

VESSL is a platform for deploying and managing AI applications. It allows you to deploy your AI applications on the cloud with a single command, and provides a web interface for managing your applications.

To deploy the FLUX.1 [schnell] Gradio app, follow the steps below:

1. Install VESSL CLI and configure your identity:

   $ pip install --upgrade vessl
   $ vessl configure

2. (Optional) Create a new project (replace ${PROJECT_NAME} with the name you want):

   $ vessl project create ${PROJECT_NAME}
   $ vessl configure -p ${PROJECT_NAME}

3. Clone this repository:

   $ git clone https://github.com/vessl-ai/examples.git
   $ cd examples/runs/flux.1-schnell

4. Deploy the application with the provided YAML file (run.yaml):

   $ vessl run create -f run.yaml

For additional information and support, please refer to the VESSL documentation.