Yosef-Ali/Gemma3-Amharic-language

Gemma 3 for Amharic language # Gemma 3 for Amharic Language Applications This project leverages the Gemma 3 model (including its vision capabilities) for developing Amharic language applications, specifically focusing on chat/audio and OCR applications. It builds upon existing work with the Walia-LLM model and the amharic-dataset-mcp tools. ## 🎯 Project Goals 1. **OCR Application**: Develop an efficient OCR system for Amharic text using Gemma 3 Vision 2. **Chat/Audio Application**: Create a conversational AI that understands and responds in Amharic 3. **Efficiency**: Leverage Gemma 3's smaller size (270M parameters) for faster training and inference 4. **Integration**: Combine with existing professional datasets and tools ## 📁 Project Structure ``` . ├── gemma3_amharic_plan.md # Detailed implementation plan ├── gemma3_amharic_demo.py # Demonstration script ├── gemma3_amharic_requirements.txt # Python dependencies ├── gemma3_amharic_setup.sh # Setup script ├── README.md # This file ├── amharic-dataset-mcp/ # MCP tools for Amharic data (separate repo) ├── amharic-llm-data/ # Walia-LLM project with professional datasets ``` ## 🚀 Quick Start 1. **Run the setup script**: ```bash ./gemma3_amharic_setup.sh ``` 2. **Activate the environment**: ```bash source gemma3_amharic_env/bin/activate ``` 3. **Install additional dependencies**: ```bash # Install protobuf (required for Walia-LLM) pip install protobuf # Install Unsloth for faster training (optional) pip install git+github.com ``` 4. **Set up HuggingFace authentication** (for Gemma models): ```bash # Get your HuggingFace token from huggingface.co huggingface-cli login # Enter your token when prompted ``` 5. **Run the demonstration**: ```bash python gemma3_amharic_demo.py ``` ## 🔐 Important Authentication Notes - **Gemma Models**: Require HuggingFace authentication with access to gated repositories - **Walia-LLM* …