README.md

metadata

title: LightOnOCR 1B Demo
emoji: 📖
colorFrom: blue
colorTo: indigo
sdk: gradio
sdk_version: 5.42.0
app_file: app_space.py
pinned: false
license: other

LightOnOCR-1B Demo

High-performance OCR application using LightOnOCR-1B model, optimized for Apple Silicon and ZeroGPU.

🚀 Performance

• GGUF Backend: ~3-4 seconds per page (M3 Max)!
• PyTorch Backend: ~40 seconds per page

Features

• 📄 PDF and image support
• 🔄 Seamless switching between GGUF and PyTorch backends (Local)
• 🎛️ Configurable resolution (scale) and token generation
• 🖥️ CLI and Gradio web interface
• 🍎 Full Metal/MPS support

Quick Start (Local)

1. Prerequisites

• Python 3.10+
• cmake and git

pip install -r requirements.txt
pip install accelerate

2. Setup GGUF (Highly Recommended)

See GGUF Setup Guide.

1. Build llama.cpp locally:

   git clone https://github.com/ggerganov/llama.cpp
   cd llama.cpp && mkdir build && cd build
   cmake .. -DGGML_METAL=ON && cmake --build . --config Release -j 8
   cd ../..
   

2. Download model:

   python download_gguf_model.py
   

3. Usage

Command Line:

# Fastest
python ocr_cli.py document.pdf --backend gguf

# High Quality
python ocr_cli.py document.pdf --backend gguf --scale 2.0

Web Interface:

python app.py

Open http://127.0.0.1:7860 and select GGUF backend.

Documentation

• GGUF Setup Guide
• Performance Optimization