Docker Environment Collection

This repository is a collection of Docker and Docker Compose-based environment configurations tailored for various development scenarios.

---

🎯 Overview

This repository contains multiple, independent Docker environments, each located in its own sub-directory. The goal is to provide isolated, reproducible, and portable setups for specific tasks.

• Reproducible: Dockerfile specifications ensure the same environment can be built every time.
• Portable: These environments can be reproduced on any machine that runs Docker.
• Efficient: Start and stop entire environments with simple docker compose commands.

---

🔧 Environments Included

1. cuda_llm_server

PyTorch & Ollama GPU Development Environment

This environment is designed for machine learning projects that require both PyTorch for development and a persistent Ollama server for LLM inference, all utilizing NVIDIA GPUs.

• Base: Ubuntu 22.04, NVIDIA CUDA Toolkit 12.1
• Key Software: Python 3.11, PyTorch (with CUDA 12.1 support), Ollama Server
• Features: The Ollama server starts automatically on container launch, and model data is saved in a persistent Docker volume.

2. python_cuda_torch

PyTorch GPU Development Environment

A simple, focused environment for machine learning development using PyTorch with NVIDIA GPU acceleration.

• Base: Ubuntu 22.04, NVIDIA CUDA Toolkit 12.1
• Key Software: Python 3.10, PyTorch (with CUDA 12.1 support)
• Features: Ideal for model training and inference scripts that require GPU access.

3. python_data_science

Data Science Development Environment (VS Code Dev Container)

A data analysis environment built on the Jupyter data science stack and optimized for use with VS Code Dev Containers.

• Base: jupyter/datascience-notebook:x86_64-ubuntu-22.04
• Key Software: Python 3.10, pandas, numpy
• Features: Designed to be opened directly in VS Code ("Reopen in Container") for a seamless data science workflow.

---

✅ General Prerequisites

For the GPU-enabled environments (cuda_llm_server and python_cuda_torch), you must have the following installed and configured on your host machine:

• An NVIDIA GPU
• An up-to-date NVIDIA Driver
• Docker
• A Docker environment configured for GPU access:
  • (For Windows) WSL 2 integration enabled in Docker Desktop
  • (For Linux) NVIDIA Container Toolkit

---

🚀 How to Use

1. Clone this repository.
2. Navigate into the sub-directory of the environment you wish to use (e.g., cd cuda_llm_server).
3. Follow the instructions in that directory's specific README.md to build and run the container.

Example (for cuda_llm_server):

# Navigate to the environment directory
cd cuda_llm_server

# Build and run the container in the background
docker compose up --build -d

# Enter the running container
docker compose exec app /bin/bash