🐳 Mini-NanoGPT Docker Deployment Guide

English · 简体中文

---

Note

The files in this directory are used for running and initializing in containers.

This project provides a complete Docker solution with automatic CUDA detection and corresponding PyTorch environment selection.

---

🚀 Quick Start

Using Docker Compose (Recommended)

# Start container in foreground
docker-compose up --build

# Stop services
docker-compose down

Using Docker Commands

# Build image
docker build -t mini-nanogpt .

# Run container (auto GPU detection)
docker run --gpus all -p 7860:7860 \
  -v $(pwd)/data:/app/data \
  -v $(pwd)/out:/app/out \
  -v $(pwd)/assets:/app/assets \
  mini-nanogpt

# Run container (CPU only mode)
docker run -p 7860:7860 \
  -v $(pwd)/data:/app/data \
  -v $(pwd)/out:/app/out \
  -v $(pwd)/assets:/app/assets \
  mini-nanogpt

---

🐛 Troubleshooting

1. GPU Not Recognized

# Check NVIDIA drivers
nvidia-smi

# Check Docker GPU support
docker run --rm --gpus all nvidia/cuda:11.8-base-ubuntu22.04 nvidia-smi

2. Port Already in Use

# Modify port mapping in docker-compose.yml
ports:
  - "8080:7860"  # Use port 8080

3. Insufficient Memory

# Check system resources
docker stats

# Limit container memory usage
docker run -m 4g mini-nanogpt

View Logs

# Docker Compose logs
docker-compose logs -f

# Docker container logs
docker logs mini-nanogpt

---

🔄 Updates and Maintenance

# Rebuild image
docker-compose build --no-cache

# Clean unused images
docker image prune

# Complete reset
docker-compose down
docker system prune -a

---

📝 Environment Variables

You can customize configuration through environment variables:

Variable	Default	Description
GRADIO_SERVER_NAME	0.0.0.0	Server bind address
GRADIO_SERVER_PORT	7860	Server port
PYTHONUNBUFFERED	1	Python output buffering
MINI_NANOGPT_ENV_TYPE	AUTO	Environment type: AUTO, CUDA, CPU

Example docker-compose.yml:

environment:
  - GRADIO_SERVER_NAME=0.0.0.0
  - GRADIO_SERVER_PORT=7860
  - PYTHONUNBUFFERED=1
  - MINI_NANOGPT_ENV_TYPE=AUTO