Skip to content

Latest commit

 

History

History
158 lines (119 loc) · 3.78 KB

DOCKER_COMPOSE.md

File metadata and controls

158 lines (119 loc) · 3.78 KB

Docker Compose files for local LoRA Pilot deployment

📁 Files Created

Core Files

  • docker-compose.yml - Standard GPU setup with all services
  • docker-compose.dev.yml - Development setup with source mounting
  • docker-compose.cpu.yml - CPU-only setup (limits via local Compose are not enforced)
  • .env.example - Environment variables template

Documentation & Tools

  • docker-compose/README.md - Comprehensive setup guide
  • docker-compose-helper.sh - Easy management script

🚀 Quick Start

1. Basic Setup

# Clone and navigate to lora-pilot directory
cd lora-pilot

# Setup workspace and environment
./docker-compose-helper.sh setup

# Start with GPU support
./docker-compose-helper.sh start

# Or start CPU-only
./docker-compose-helper.sh start cpu

2. Access Services

🛠️ Management Commands

# Start services
./docker-compose-helper.sh start [standard|dev|cpu]

# Stop services  
./docker-compose-helper.sh stop [standard|dev|cpu]

# View logs
./docker-compose-helper.sh logs [standard|dev|cpu]

# Access container shell
./docker-compose-helper.sh shell [standard|dev|cpu]

# Check status
./docker-compose-helper.sh status [standard|dev|cpu]

# Update image
./docker-compose-helper.sh update

# Setup workspace
./docker-compose-helper.sh setup

⚙️ Configuration Options

Environment Variables (.env)

# Timezone
TZ=America/New_York

# Hugging Face token
HF_TOKEN=your_token_here

# Custom ports
JUPYTER_PORT=8888
COMFY_PORT=5555
AI_TOOLKIT_PORT=8675
# ... etc

Volume Mounting

volumes:
  # Main workspace
  - ./workspace:/workspace
  
  # Custom models
  - ./my-models:/workspace/models
  
  # Custom datasets
  - ./my-datasets:/workspace/datasets

🎯 Use Cases

Standard Setup (docker-compose.yml)

  • ✅ Full GPU support
  • ✅ All services exposed
  • ✅ Health checks
  • ✅ Production-ready

Development Setup (docker-compose.dev.yml)

  • ✅ Source code mounting
  • ✅ Debug mode enabled
  • ✅ Interactive shell access
  • ✅ Live reloading

CPU-Only Setup (docker-compose.cpu.yml)

  • ✅ No GPU requirements
  • ✅ Minimal ports
  • ✅ Optimized for CPU inference
  • ⚠️ Resource limits are not enforced by local Docker Compose (use Docker runtime limits if needed)

🐛 Troubleshooting

Common Issues

  1. GPU not detected: Check NVIDIA Docker installation
  2. Port conflicts: Change ports in compose file
  3. Permission errors: Fix workspace permissions
  4. Memory issues: Use Docker runtime limits (--memory, --cpus) or Docker Desktop settings

Debug Commands

# Check container status
docker-compose ps

# View logs
./docker-compose-helper.sh logs

# Access shell
./docker-compose-helper.sh shell

# Check GPU access
docker run --rm --gpus all nvidia/cuda:12.4.1-runtime-ubuntu22.04 nvidia-smi

📚 Documentation

See docker-compose/README.md for comprehensive documentation including:

  • Prerequisites and installation
  • Customization options
  • Troubleshooting guide
  • Production tips
  • Additional resources

🎉 Benefits

  • Easy Setup: One-command deployment
  • Multiple Configurations: GPU, CPU, Development modes
  • Persistent Storage: Workspace data preserved
  • Health Monitoring: Built-in health checks
  • Resource Management: Use Docker runtime limits if needed
  • Development Friendly: Source mounting and debug mode

Users can now easily run LoRA Pilot locally with professional Docker Compose setup! 🚀✨