Available Docker Images

When renting a GPU server on Clore.ai, you can choose from pre-configured Docker images or use your own.

Pre-installed Images

General Purpose

Image	Description	Included
Ubuntu Jupyter	Clean Ubuntu with Jupyter Notebook	Python, CUDA, Jupyter
PyTorch	Ready for deep learning	PyTorch, CUDA, cuDNN
TensorFlow	Google's ML framework	TensorFlow, CUDA, Keras

AI/ML Specific

Image	Description	Use Case
Stable Diffusion WebUI	AUTOMATIC1111 WebUI	Image generation
ComfyUI	Node-based SD interface	Advanced image workflows
Text Generation WebUI	Oobabooga's interface	LLM inference
Ollama	Run LLMs locally	LLM deployment
vLLM	High-performance LLM serving	Production LLM

Mining

Image	Description
Clore Mining	Pre-configured for mining
HiveOS Compatible	For HiveOS integration

Selecting an Image

1. Go to Marketplace

2. Find a server

3. Click Rent

4. In the order form, select Docker Image from dropdown

5. Configure ports if needed

6. Submit order

Using Custom Docker Images

You can specify any public Docker image from Docker Hub or other registries.

Format

registry/image:tag

Examples

pytorch/pytorch:2.1.0-cuda12.1-cudnn8-runtime
nvidia/cuda:12.2.0-base-ubuntu22.04
huggingface/transformers-pytorch-gpu:latest

Requirements for Custom Images

• Must be publicly accessible (or use auth)

• Should be NVIDIA GPU compatible

• Base on CUDA-enabled images for GPU access

Port Configuration

When creating an order, you can expose ports for your applications:

Port	Common Use
22	SSH (usually pre-configured)
8888	Jupyter Notebook
7860	Gradio apps (SD WebUI, etc.)
3000	Web applications
8080	HTTP services

Setting Custom Ports

In the order form, specify ports to expose:

8888,7860,3000

Environment Variables

You can pass environment variables to your container:

HF_TOKEN=your_token
MODEL_NAME=llama-7b

Persistent Storage

• Data in /workspace is typically persistent during the rental

• Other directories may be reset on container restart

• Always backup important data externally

Best Practices

1. Choose appropriate image - Use pre-configured images when possible

2. Check CUDA version - Ensure compatibility with your workload

3. Expose only needed ports - For security

4. Use environment variables - For configuration, not secrets

5. Test locally first - Verify your custom image works before renting

Troubleshooting

Image won't start

• Check if image exists and is public

• Verify CUDA/GPU compatibility

• Check container logs via SSH

GPU not accessible in container

• Ensure image is NVIDIA-compatible

• Check NVIDIA drivers: nvidia-smi

• Verify CUDA installation: nvcc --version

Can't access exposed ports

• Wait for container to fully start

• Check if service is running inside container

• Verify firewall/security settings