doc-page-extractor

December 11, 2025 · View on GitHub

Document page extraction tool powered by DeepSeek-OCR.

Installation

⚠️ Important: This package requires PyTorch with CUDA support (GPU Required). PyTorch is NOT automatically installed - you must install it manually first.

Step 1: Install PyTorch with CUDA

Choose the command that matches your CUDA version:

# For CUDA 12.1 (recommended for most users)
pip install torch torchvision --index-url https://download.pytorch.org/whl/cu121

# For CUDA 11.8
pip install torch torchvision --index-url https://download.pytorch.org/whl/cu118

# For CUDA 12.6
pip install torch torchvision --index-url https://download.pytorch.org/whl/cu126

💡 Don't know your CUDA version? Run nvidia-smi to check, or just try CUDA 12.1 (works with most recent drivers).

Step 2: Install doc-page-extractor

pip install doc-page-extractor

Verify Installation

Check if everything is working:

python -c "import doc_page_extractor; import torch; print('✓ Installation successful!'); print('✓ CUDA available:', torch.cuda.is_available())"

Expected output:

✓ Installation successful!
✓ CUDA available: True

If CUDA shows False, see the troubleshooting section below.

Usage

from doc_page_extractor import PageExtractor

# Your code here

Troubleshooting

"PyTorch is required but not installed!"

Install PyTorch first:

pip install torch torchvision --index-url https://download.pytorch.org/whl/cu121

"CUDA is not available!"

Check your GPU driver:

nvidia-smi

If the command fails, you need to install NVIDIA drivers:

If it succeeds, you might have CPU-only PyTorch. Reinstall with CUDA:

pip uninstall torch torchvision
pip install torch torchvision --index-url https://download.pytorch.org/whl/cu121

Requirements

  • Python >= 3.10, < 3.14
  • NVIDIA GPU with CUDA 11.8 or 12.1 support (Required)
  • Sufficient GPU memory (recommended: 4GB+ VRAM)

Dependencies & Licenses

This project is licensed under the MIT License. It depends on the DeepSeek-OCR model which uses easydict (LGPLv3) for configuration management.

Development

For contributors and developers, see Development Guide for:

  • Running tests
  • Running lint checks
  • Building the package