​
Endpoint for NVIDIA CCluster

This guide helps you deploy a FLUX endpoint on NVIDIA CCluster using a pre-built Docker image or by building and pushing your own.

​
Docker image

  • Use the pre-built image: vagias/base-api:v1.0
  • Alternatively, build your own image locally and push it to Docker Hub.

​
Building the image

  • For macOS
To build and push the image using buildx (for multi-platform support): 
docker buildx build --platform linux/amd64,linux/arm64 -t {docker-username}/flux-centml:v1.0 --push .
  • For Linux
To build and push the image: 
docker build -t {docker-username}/flux-centml:v1.0 --push .

​
Deploying on NVIDIA CCluster

  1. Log in to NVIDIA CCluster Access the NVIDIA CCluster dashboard using the console URL provided to your organization.
  2. Navigate to General Inference From the Home page, go to General Inference to set up the deployment.
  3. Fill in Deployment Details Provide the necessary details, including:
  • Docker image name
username/base-api # replace with your image
  • Tag (e.g., v1.0)
v1.0
  • Port number
8000
  • Health check path
/docs
  • Entry point arguments
python main.py
  • Your HF_TOKEN (Hugging Face API token) for authentication Example:
  1. Select Resource Size Choose the resource size based on your image generation speed requirements:
  • Small
  • Medium
  • Large
  1. Click Deploy
After deployment, you will receive an endpoint URL. Interacting with the Endpoint Once deployed, you can interact with the endpoint via:
  • curl commands
  • The included apps and examples
Use the provided URL to send requests and generate images with FLUX. This version adds clarity and a more polished flow. Let me know if you’d like to include more details or examples for interacting with the endpoint!