Skip to content

Latest commit

 

History

History
392 lines (277 loc) · 7.64 KB

installation.md

File metadata and controls

392 lines (277 loc) · 7.64 KB

Installation Guide

This guide covers all installation methods for Nano Qwen3 Serving, from basic pip installation to advanced development setups.

🎯 System Requirements

Minimum Requirements

  • Operating System: macOS 12.3+ (Monterey) or later
  • Hardware: Apple Silicon (M1, M2, M3, or M1 Pro/Max/Ultra)
  • Memory: 8GB RAM (16GB+ recommended)
  • Storage: 5GB free space for models
  • Python: 3.8, 3.9, 3.10, 3.11, or 3.12

Recommended Requirements

  • Memory: 16GB+ RAM for larger models
  • Storage: 20GB+ free space for multiple models
  • Python: 3.11 or 3.12
  • Network: Stable internet connection for model downloads

🚀 Installation Methods

Method 1: PyPI Installation (Recommended)

The easiest way to install Nano Qwen3 Serving:

pip install nano-qwen3-serving

Verify installation:

python -c "import nano_qwen3_serving; print('Installation successful!')"

Method 2: Source Installation

For development or custom modifications:

# Clone the repository
git clone https://github.com/hsliuustc/nano-qwen3-serving.git
cd nano-qwen3-serving

# Install in development mode
pip install -e .

# Or install with all dependencies
pip install -e ".[dev]"

Method 3: Conda Installation

Using conda for environment management:

# Create a new conda environment
conda create -n nano-qwen3 python=3.11
conda activate nano-qwen3

# Install the package
pip install nano-qwen3-serving

Method 4: Docker Installation

For containerized deployment:

# Pull the Docker image
docker pull hsliuustc/nano-qwen3-serving:latest

# Run the container
docker run -p 8000:8000 hsliuustc/nano-qwen3-serving:latest

📦 Dependencies

Core Dependencies

The following packages are automatically installed:

  • torch: PyTorch with MPS support
  • transformers: Hugging Face transformers library
  • fastapi: Web framework for API
  • uvicorn: ASGI server
  • pydantic: Data validation
  • loguru: Logging framework

Optional Dependencies

Install additional features:

# Development dependencies
pip install nano-qwen3-serving[dev]

# Testing dependencies
pip install nano-qwen3-serving[test]

# All dependencies
pip install nano-qwen3-serving[all]

🔧 Environment Setup

1. Python Environment

Using venv (recommended):

# Create virtual environment
python -m venv nano-qwen3-env

# Activate environment
source nano-qwen3-env/bin/activate  # macOS/Linux
# nano-qwen3-env\Scripts\activate  # Windows

# Install package
pip install nano-qwen3-serving

Using conda:

# Create conda environment
conda create -n nano-qwen3 python=3.11

# Activate environment
conda activate nano-qwen3

# Install package
pip install nano-qwen3-serving

2. Environment Variables

Set optional environment variables:

# Server configuration
export NANO_QWEN3_PORT=8000
export NANO_QWEN3_HOST=127.0.0.1
export NANO_QWEN3_MODEL=Qwen/Qwen3-0.6B
export NANO_QWEN3_DEVICE=mps

# Logging
export NANO_QWEN3_LOG_LEVEL=info

# Hugging Face
export HUGGING_FACE_HUB_TOKEN=your_token_here
export HF_HOME=./models  # Custom model cache directory

3. Model Cache Configuration

Configure where models are stored:

# Set Hugging Face cache directory
export HF_HOME=./models

# Or use Python
import os
os.environ['HF_HOME'] = './models'

🍎 Apple Silicon Setup

1. Verify Apple Silicon

Check if you have Apple Silicon:

# Check processor architecture
uname -m
# Should return: arm64

# Check processor type
sysctl -n machdep.cpu.brand_string
# Should contain: Apple M1, M2, or M3

2. Install PyTorch with MPS Support

# Install PyTorch with MPS support
pip install torch torchvision torchaudio

# Verify MPS availability
python -c "import torch; print(f'MPS available: {torch.backends.mps.is_available()}')"

3. Test MPS Functionality

import torch

# Check MPS availability
if torch.backends.mps.is_available():
    device = torch.device("mps")
    print("MPS is available!")
else:
    device = torch.device("cpu")
    print("MPS not available, using CPU")

🔍 Verification

1. Basic Verification

Test the installation:

# Check if package is installed
python -c "import nano_qwen3_serving; print('Package installed successfully')"

# Check version
python -c "import nano_qwen3_serving; print(nano_qwen3_serving.__version__)"

2. Model Download Test

Test model downloading:

from transformers import AutoTokenizer, AutoModelForCausalLM

# Test small model download
model_name = "Qwen/Qwen3-0.6B"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name)

print("Model downloaded successfully!")

3. Server Test

Test the server:

# Start server in background
python -m nano_qwen3_serving --port 8001 &

# Test health endpoint
curl http://localhost:8001/health

# Stop server
pkill -f "nano_qwen3_serving"

🛠️ Development Setup

1. Clone Repository

git clone https://github.com/hsliuustc/nano-qwen3-serving.git
cd nano-qwen3-serving

2. Install Development Dependencies

# Install in development mode
pip install -e ".[dev]"

# Or install manually
pip install -e .
pip install pytest pytest-asyncio black isort mypy

3. Setup Pre-commit Hooks

# Install pre-commit
pip install pre-commit

# Install hooks
pre-commit install

4. Run Tests

# Run all tests
pytest

# Run with coverage
pytest --cov=nano_qwen3_serving

# Run specific test file
pytest tests/test_basic.py

🐳 Docker Setup

1. Build Docker Image

# Build from source
docker build -t nano-qwen3-serving .

# Or pull from registry
docker pull hsliuustc/nano-qwen3-serving:latest

2. Run Container

# Basic run
docker run -p 8000:8000 nano-qwen3-serving

# With custom model
docker run -p 8000:8000 -e NANO_QWEN3_MODEL=Qwen/Qwen3-1.5B nano-qwen3-serving

# With volume for model cache
docker run -p 8000:8000 -v ./models:/app/models nano-qwen3-serving

3. Docker Compose

Create docker-compose.yml:

version: '3.8'
services:
  nano-qwen3:
    image: hsliuustc/nano-qwen3-serving:latest
    ports:
      - "8000:8000"
    environment:
      - NANO_QWEN3_MODEL=Qwen/Qwen3-0.6B
      - NANO_QWEN3_DEVICE=mps
    volumes:
      - ./models:/app/models
    restart: unless-stopped

Run with:

docker-compose up -d

🔧 Troubleshooting Installation

Common Issues

  1. Permission Errors:

    # Use user installation
pip install --user nano-qwen3-serving

  2. PyTorch MPS Issues:

    # Reinstall PyTorch
pip uninstall torch torchvision torchaudio
pip install torch torchvision torchaudio

  3. Model Download Issues:

    # Set Hugging Face token
export HUGGING_FACE_HUB_TOKEN=your_token

# Or use mirror
export HF_ENDPOINT=https://hf-mirror.com

  4. Memory Issues:

    # Use smaller model
python -m nano_qwen3_serving --model Qwen/Qwen3-0.6B

Getting Help

📚 Next Steps

After installation:

  1. Quick Start: Get up and running in minutes
  2. Configuration: Configure the server
  3. API Reference: Learn the API
  4. Examples: See usage examples

🎉 Installation complete! You're ready to start using Nano Qwen3 Serving!