Python Humpback Docker Images

This repository contains Docker images for a Python-based application, with support for both a base environment and an environment configured for CUDA/cuDNN.

Images

Base Image

The base image provides a minimal Python environment with essential tools and dependencies for running the application.

cuDNN Image

The cudnn image builds on the base image and includes NVIDIA CUDA and cuDNN libraries for GPU acceleration.

Features

• Base Image:

  • Python 3.12 (slim variant)
  • Virtual environment setup
  • Common build tools (e.g., build-essential, curl, git)
  • Pre-installed tini as an init system
  • Configured entrypoint script for running Python applications

• cuDNN Image:

  • NVIDIA CUDA/cuDNN support
  • Configured for GPU passthrough with NVIDIA_VISIBLE_DEVICES and NVIDIA_DRIVER_CAPABILITIES

Usage

Pull the Images

docker pull ghcr.io/ai-dope/python-humpback-base:latest

docker pull ghcr.io/ai-dope/python-humpback-cudnn:latest

Run Your App

Humpback runs your application through a volume mount on the /app/ directory and by default will attempt to execute python /app/app.py, but this can be overridden with your own CMD [] statement or post docker args. You can optionally persist the python cache at /root/.cache/ to wherever you'd like.

Example

cd repos/my_sweet_app/
docker container run -v /$(pwd)$:/app --runtime=nvidia ghcr.io/ai-dope/python-humpback-cudnn:latest python /app/main.py