Running DeepSeek Prover V2 7B on Ubuntu involves a detailed process that includes setting up your environment, preparing GPU infrastructure, installing dependencies, and configuring the model for local or cloud-based use.

This guide walks you through all essential steps to get DeepSeek Prover V2 7B up and running on an Ubuntu system.

What is DeepSeek Prover V2 7B?

DeepSeek Prover V2 7B is a powerful large language model (LLM) built for formal theorem proving in Lean 4—a programming language and interactive proof assistant. With 7 billion parameters, it is designed to assist in complex mathematical verification by generating and validating proofs.

To run efficiently, the model benefits from GPU acceleration, particularly on systems equipped with NVIDIA GPUs and CUDA support.

System Requirements

Ensure your system meets these minimum specifications before installation:

• OS: Ubuntu 20.04 or later (Ubuntu 22.04 or 24.04 recommended)
• CPU: 4-core minimum
• RAM: At least 16 GB (32 GB preferred)
• Storage: Minimum 12 GB free space
• GPU: NVIDIA with CUDA support (16 GB+ VRAM recommended)
• Python: Version 3.10 or higher

Step-by-Step Installation Guide

1. Update Your Ubuntu System

sudo apt update && sudo apt upgrade -y

Install essential packages:

sudo apt install -y python3 python3-pip python3.10 python3.10-distutils git software-properties-common

Add Python 3.10 via deadsnakes PPA if needed:

sudo add-apt-repository -y ppa:deadsnakes/ppa
sudo apt update
sudo apt install -y python3.10 python3.10-distutils

Verify Python version:

python3 --version

2. Set Up NVIDIA Drivers and CUDA

Check for GPU driver:

nvidia-smi

Install or update if necessary:

sudo apt install -y nvidia-driver-525
sudo reboot

Install CUDA toolkit (example for CUDA 11.8):

sudo apt install -y cuda-toolkit-11-8

3. Install PyTorch with CUDA

Install PyTorch with CUDA support:

pip3 install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu118

Check CUDA availability:

python3 -c "import torch; print(torch.cuda.is_available())"

Should return True.

4. Clone DeepSeek Prover V2 Repository

Clone from GitHub:

git clone https://github.com/deepseek-ai/DeepSeek-Prover-V2.git
cd DeepSeek-Prover-V2

Or pull the model weights from Hugging Face:

git lfs install
git clone https://huggingface.co/deepseek-ai/DeepSeek-Prover-V2-7B

5. Install Required Python Packages

Inside the project folder, install dependencies:

pip3 install -r requirements.txt

If the file is missing, install manually:

pip3 install transformers torch tokenizers sentencepiece

Install lean4 if needed for theorem proving tasks.

6. Run the DeepSeek Prover V2 7B Model

Run the model using the provided script:

python3 run_deepseek_prover.py --model-path ./DeepSeek-Prover-V2-7B --device cuda

Adjust the script name and parameters based on repository documentation.

7. Optional: Use NodeShift for Cloud GPUs

If your local GPU is underpowered:

1. Sign up at NodeShift
2. Launch a GPU-enabled VM (e.g., A100 or H100)
3. Deploy a CUDA-enabled image
4. SSH into the instance and repeat the above steps

8. Alternative Setup: Using Ollama

Ollama offers a simplified interface for running models locally:

Install Ollama:

curl -fsSL https://ollama.com/install.sh | sh

Check installation:

ollama --version

Pull the model:

ollama pull deepseek-r1:7b

Run the model:

ollama run deepseek-r1:7b

This approach reduces setup complexity significantly.

Troubleshooting and Optimization Tips

• Out of Memory (OOM) Errors: Use a smaller model or quantization techniques.
• Python Conflicts: Use venv or conda for isolated environments.
• GPU Limits: Multi-GPU setups may be required for larger workloads.
• Swap File (if RAM-limited):

sudo fallocate -l 8G /swapfile
sudo chmod 600 /swapfile
sudo mkswap /swapfile
sudo swapon /swapfile

• Security: Use SSH keys and keep dependencies updated on cloud VMs.

Conclusion

Deploying DeepSeek Prover V2 7B on Ubuntu involves:

• Installing Python 3.10+ and GPU-compatible PyTorch
• Setting up NVIDIA drivers and CUDA
• Downloading the model and its dependencies
• Running it locally or on GPU-powered cloud platforms like NodeShift
• Optionally using Ollama for a simplified experience

With the right setup, DeepSeek Prover V2 7B empowers formal proof generation at scale, making it a valuable tool for anyone working in AI-assisted theorem proving.

References

1. Run DeepSeek Janus-Pro 7B on Mac: A Comprehensive Guide Using ComfyUI
2. Run DeepSeek Janus-Pro 7B on Mac: Step-by-Step Guide
3. Run DeepSeek Janus-Pro 7B on Windows: A Complete Installation Guide
4. Running DeepSeek Prover V2 7B on macOS: A Comprehensive Guide
5. Running DeepSeek Prover V2 7B on Windows: A Complete Setup Guide