Skip to content

Latest commit

 

History

History
95 lines (75 loc) · 2.37 KB

windows.md

File metadata and controls

95 lines (75 loc) · 2.37 KB

Windows Setup Guide

This guide covers setting up DeepSeek R1 Local on Windows, both with and without GPU support.

Note: Commands may need to be run with sudo if you are not in the docker group.

Setup Options

Option 1: CPU-Only (Native Windows)

  1. Install Docker Desktop for Windows:

    • Download from Docker's official website
    • Run the installer and follow the prompts
    • You do not need to enable WSL2 integration for CPU-only usage

  2. Start the containers:

docker compose up -d

Option 2: GPU Support (WSL2)

Prerequisites

  • NVIDIA GPU with updated drivers
  • Windows 10/11 with WSL2 enabled
  • Ubuntu installed in WSL2

Setup Steps

  1. Enable WSL2:
wsl --install
  1. Install Make utility:
sudo apt-get install make
  1. Install Docker and Docker Compose in WSL2:
make install-docker
  1. Install NVIDIA Container Toolkit in WSL2:
make setup-gpu
If successful, you should see something like the following, if not, see [Troubleshooting](#troubleshooting):
![GPU Test](../images/gpu-test.png)
  1. Start the containers with GPU support:
docker compose -f docker-compose.yml -f docker-compose.gpu.yml -f docker-compose.windows.yml up -d

Troubleshooting

CPU-Only (Docker Desktop)

  1. If containers fail to start:
    • Check Docker Desktop is running
    • Ensure you have enough memory allocated in Docker Desktop settings
    • Try restarting Docker Desktop

GPU Support (WSL2)

  1. If GPU is not detected:

    • Ensure you have the latest NVIDIA drivers installed in Windows
    • Check if WSL2 can see the GPU: 
      nvidia-smi
    • If nvidia-smi fails, try updating WSL2: 
      wsl --update

  2. If Docker can't access the GPU:

    • Check if the Docker daemon is running: 
      sudo systemctl status docker
    • Verify NVIDIA Container Toolkit installation: 
      nvidia-ctk --version
    • Restart the Docker daemon: 
      sudo systemctl restart docker

Performance Notes

  • WSL2 generally provides better performance than native Windows, even for CPU-only workloads
  • The first startup will be slower due to model download (1.1GB for default model)
  • Models are stored in the ollama-models Docker volume and persist between restarts