phygrid/cuda-base repository overview

⁠Phygrid CUDA Base Image

A multi-architecture Docker base image optimized for AI inference services, providing common system dependencies and Python packages for CUDA-accelerated applications. Supports both Intel/AMD x64 systems and ARM64 NVIDIA Jetson devices.

⁠🚀 Quick Start

# Pull the latest image
docker pull phygrid/cuda-base:latest

# Use as base image in your Dockerfile
FROM phygrid/cuda-base:1.0.0

⁠📋 What's Included

⁠System Dependencies

• Build tools: build-essential, cmake, git, wget, curl
• Audio processing: libasound2-dev, portaudio19-dev, libsndfile1, ffmpeg
• Image processing: libgl1, libglib2.0-0, OpenCV dependencies
• Security: patchelf for executable stack fixes
• Networking: ca-certificates and utilities

⁠Python Environment

• Base: Python 3.11 with optimized pip, setuptools, wheel
• Web frameworks: FastAPI 0.104.1, Uvicorn, Pydantic 2.5.0
• Core libraries: NumPy 1.24.4, Pillow 10.1.0, Requests 2.31.0
• Utilities: aiofiles, python-dotenv, python-multipart

⁠Container Features

• Security: Non-root appuser with proper permissions
• Structure: Pre-created /app/{cache,models,data,logs} directories
• Health check: Built-in health check endpoint
• Multi-arch: AMD64 (Intel/AMD) and ARM64 (NVIDIA Jetson) support
• CUDA: GPU acceleration for NVIDIA Blackwell and earlier architectures
• Port: Exposes port 8000 (customizable)

⁠🐳 Docker Hub

Repository: phygrid/cuda-base⁠

⁠Available Tags

• latest - Latest stable release
• 1.0.0, 1.0.1, etc. - Specific semantic versions
• Multi-architecture support: linux/amd64, linux/arm64

⁠📦 Usage Examples

⁠As Base Image

FROM phygrid/cuda-base:1.0.0

# Copy your application
COPY . /app/

# Install additional dependencies
RUN pip install -r requirements.txt

# Override default command
CMD ["python", "main.py"]

⁠Development Environment

# Run interactive container
docker run -it --rm \
  -v $(pwd):/app/workspace \
  -p 8000:8000 \
  phygrid/cuda-base:latest \
  bash

⁠Production Deployment

# Run with GPU support (Intel/AMD systems)
docker run -d \
  --name my-ai-service \
  --gpus all \
  -p 8000:8000 \
  -v /data:/app/data \
  phygrid/cuda-base:latest

# Run on NVIDIA Jetson devices
docker run -d \
  --name my-ai-service \
  --runtime nvidia \
  --gpus all \
  -p 8000:8000 \
  -v /data:/app/data \
  phygrid/cuda-base:latest

⁠🏗️ Building from Source

# Clone repository
git clone https://github.com/phygrid/cuda-base.git
cd cuda-base

# Build image
docker build -t phygrid/cuda-base:custom .

# Build for multiple architectures
docker buildx build --platform linux/amd64,linux/arm64 -t phygrid/cuda-base:custom .

⁠🔄 Versioning

This project uses automated semantic versioning:

• Automatic: Patch versions increment on main branch changes
• Manual: Edit VERSION file for major/minor bumps
• Tags: Git tags created automatically (e.g., v1.0.0)

See DOCKER_DEPLOYMENT.md⁠ for detailed deployment workflow.

⁠🧪 Health Check

The image includes a built-in health check:

# Test health endpoint
docker run --rm phygrid/cuda-base:latest python /app/health_check.py

⁠🤝 Contributing

We welcome contributions! Please:

1. Fork the repository
2. Create a feature branch (git checkout -b feature/amazing-feature)
3. Commit changes (git commit -m 'Add amazing feature')
4. Push to branch (git push origin feature/amazing-feature)
5. Open a Pull Request

⁠Development Setup

# Clone and setup
git clone https://github.com/phygrid/cuda-base.git
cd cuda-base

# Test build locally
docker build -t phygrid/cuda-base:test .
docker run --rm phygrid/cuda-base:test python /app/health_check.py

⁠📄 License

This project is licensed under the MIT License - see the LICENSE⁠ file for details.

⁠🔧 Environment Variables

⁠🏷️ Labels

The image includes standard OCI labels:

LABEL org.opencontainers.image.title="Phygrid CUDA Base"
LABEL org.opencontainers.image.description="Common CUDA base image for AI inference services"
LABEL org.opencontainers.image.vendor="Phygrid"
LABEL org.opencontainers.image.version="1.0.0"

⁠🆘 Support

• Issues: GitHub Issues⁠
• Discussions: GitHub Discussions⁠
• Docker Hub: phygrid/cuda-base⁠

⁠📈 Metrics

• Image size: ~800MB compressed (AMD64), ~1.2GB (ARM64 with CUDA)
• Build time: ~5-10 minutes (with cache)
• Architectures: AMD64 (Intel/AMD), ARM64 (NVIDIA Jetson)
• Python version: 3.12 (Ubuntu 24.04 default)
• Base OS: Ubuntu 24.04 + CUDA 13.0 (unified for both architectures)
• CUDA version: 13.0.0 with compatibility layer for edge deployment