Running DeepSeek Models Locally with Ollama: Complete Guide

# Using Windows Package Manager (winget)
winget install Ollama.Ollama

# Or download and run the installer from:
# https://ollama.ai/download/windows

# Using Homebrew
brew install ollama

# Or using curl
curl -fsSL https://ollama.ai/install.sh | sh

# Using curl
curl -fsSL https://ollama.ai/install.sh | sh

# For Ubuntu/Debian
curl -fsSL https://ollama.ai/install.sh | sudo bash

# List available DeepSeek models
ollama list

# Pull DeepSeek models
ollama pull deepseek-coder:6.7b
ollama pull deepseek-coder:33b
ollama pull deepseek-llm:7b

# modelfile
FROM deepseek-coder:6.7b
PARAMETER temperature 0.7
PARAMETER top_p 0.9
PARAMETER num_ctx 8192

# System prompt configuration
SYSTEM """You are an AI programming assistant. Write code that is:
- Well-documented
- Efficient
- Following best practices
"""

# Build custom model from modelfile
ollama create deepseek-custom -f modelfile

# Verify model creation
ollama list

# Start interactive session
ollama run deepseek-coder:6.7b

# Or use single command
echo "Write a Python function to calculate fibonacci numbers" | ollama run deepseek-coder:6.7b

# Example response from DeepSeek Coder
def fibonacci(n):
    """
    Calculate the nth Fibonacci number using dynamic programming.
    
    Args:
        n (int): The position in the Fibonacci sequence (0-based)
        
    Returns:
        int: The nth Fibonacci number
    """
    if n <= 1:
        return n
        
    fib = [0] * (n + 1)
    fib[1] = 1
    
    for i in range(2, n + 1):
        fib[i] = fib[i-1] + fib[i-2]
        
    return fib[n]

# Using Python with Ollama API
import requests

def query_deepseek(prompt):
    response = requests.post('http://localhost:11434/api/generate',
        json={
            'model': 'deepseek-coder:6.7b',
            'prompt': prompt,
            'stream': False
        })
    return response.json()['response']

# Example usage
code = query_deepseek("Write a function to sort a list in Python")
print(code)

# Check GPU availability
nvidia-smi

# Enable GPU in Ollama
export OLLAMA_CUDA=1
export OLLAMA_GPU_LAYERS=35

# Configure memory limits
export OLLAMA_HOST=0.0.0.0
export OLLAMA_MODELS=/path/to/models
export OLLAMA_MEMORY_LIMIT=16GB

# Modelfile with optimized parameters
FROM deepseek-coder:6.7b
PARAMETER num_ctx 4096  # Adjust based on your RAM
PARAMETER num_gpu 1
PARAMETER num_thread 8

# Check Ollama status
ollama serve status

# View logs
tail -f ~/.ollama/logs/ollama.log

# Reset Ollama
ollama kill
ollama serve

# Monitor memory usage
ps aux | grep ollama
nvidia-smi -l 1  # GPU memory monitoring

# Clear model cache
rm -rf ~/.ollama/models/*
ollama pull deepseek-coder:6.7b  # Re-download model

# Regular maintenance
ollama list  # Check installed models
ollama rm unused-model  # Remove unused models
ollama pull model:latest  # Update models

# Docker compose configuration
version: '3.8'
services:
  ollama:
    image: ollama/ollama
    container_name: ollama-deepseek
    ports:
      - "11434:11434"
    volumes:
      - ./models:/root/.ollama/models
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: [gpu]

# Secure API access
export OLLAMA_HOST=localhost  # Restrict to local access
export OLLAMA_ORIGINS=https://your-app.com  # CORS configuration

# Use environment variables for sensitive settings
export OLLAMA_API_KEY="your-api-key"