deepseek-r1 70b hardware requirements

Hardware Requirements for DeepSeek-R1 70B

Component	Requirement
GPU	Multi-GPU setup with at least 32 GB VRAM per GPU (e.g., NVIDIA A100 80GB x16)
RAM	Minimum 64 GB system memory
CPU	High-performance multi-core processor (e.g., AMD EPYC or Intel Xeon)

How to Install DeepSeek-R1 70B Locally on Windows

1. Install Windows Subsystem for Linux (WSL):

   • Ensure WSL is enabled on your Windows system.
   • Install a Linux distribution from the Microsoft Store (e.g., Ubuntu).

2. Set Up the Environment:

   • Open the WSL terminal.
   • Update package lists:

    sudo apt-get update
   

• Install necessary dependencies:
  

   sudo apt-get install -y git-lfs python3-pip
  

1. Clone the DeepSeek-R1 Repository:

   • Install Git Large File Storage (Git LFS):

    git lfs install
   

• Clone the repository:
  

   git clone https://huggingface.co/deepseek-ai/DeepSeek-R1
   cd DeepSeek-R1
  

1. Set Up a Python Virtual Environment:

   • Install virtualenv:

    pip3 install virtualenv
   

• Create and activate the virtual environment:
  

   virtualenv venv
   source venv/bin/activate
  

1. Install Python Dependencies:

   • Within the virtual environment, install required packages:

    pip install -r requirements.txt
   

2. Configure GPU Support:

   • Ensure your GPU drivers are up-to-date on Windows.
   • Install CUDA and cuDNN compatible with your GPU.
   • Verify that the GPU is accessible within WSL.

3. Run the Model:

   • Execute the model inference script:

    python run_inference.py --model_path ./DeepSeek-R1
   

Note: Due to the extensive hardware requirements, it is recommended to start with a smaller version of the model before scaling up to the full 70B parameter model.