🌟 Demo

Demo.mp4

Multilingual Translation Model

A scalable and efficient multilingual translation system powered by fine-tuned Large Language Models, featuring real-time translation capabilities across multiple language pairs.

Table of Contents

• Flowchart: Development Process
• Technical Implementation
• Performance Metrics
• Tech Stack
• Getting Started
• Project Structure
• API Documentation
• Contributing
• License

Here’s the updated section with the live link separated below the image and set to open in a new tab:

---

🛠️ Flowchart

Live Link

Live Deployment on Google Cloud Platform

Note: The live deployment is hosted on Google Cloud Platform. Please allow up to 2 minutes for cold start if the service is inactive.

---

This ensures the image and link are separate, and the link opens in a new tab when clicked. Let me know if you need further adjustments! 🚀

---

🛠️ Technical Implementation

Model Architecture

• Fine-tuned Large Language Model using LoRA (Low-Rank Adaptation)
• Quantization techniques for model optimization
• Automated validation and preprocessing pipelines

Backend

• Flask-based REST API server
• Scikit-learn for data preprocessing and validation
• Docker containerization for consistent deployment
• CI/CD pipeline integration for automated deployment

Frontend

• Responsive web interface
• Real-time translation updates
• Concurrent translation support

📊 Performance Metrics

• Translation Accuracy: 32% BLEU score
• Training Efficiency: 20% reduction in training cycle time
• Inference Latency: Sub-400ms response time
• Scalability: Handles concurrent translation requests

💻 Tech Stack

🤖 ML/Deep Learning

🎨 Frontend & 🔧 Backend

☁️ Cloud/DevOps

🚀 Getting Started

Prerequisites

# Clone the repository
git clone https://github.com/yourusername/multilingual-translation.git

# Install dependencies
pip install -r requirements.txt

Running Locally

# Start the Flask server
python app.py

# Access the web interface
open http://localhost:5000

Docker Deployment

# Build Docker image
docker build -t translation-model .

# Run container (map port 5000 to 8080)
docker run -p 8080:5000 translation-model

📁 Project Structure

.
├── Backend/
│   ├── ml/
│   │   ├── models/                                         # Saved model files for local/production use
│   │   └── notebooks/                                      # Jupyter notebooks for experiments
│   │       ├── EDA.ipynb                                   # Exploratory Data Analysis
│   │       ├── NLP pipeline.ipynb                          # Data preprocessing pipeline
│   │       ├── M2M100.ipynb                                # Base M2M100 implementation
│   │       ├── M2M100(Lora).ipynb                          # LoRA fine-tuning
│   │       ├── 4bit FB(M2M100)+lora.ipynb                  # 4-bit quantization
│   │       ├── 4bit FB(M2m100)+lora+Greek+Hindi.ipynb      # Multi-language support
│   │       └── Bleu_score_BFT.ipynb                        # Performance evaluation
│   ├── api/                                                # REST API endpoints
│   ├── dataset/                                            # Training and evaluation datasets
│   └── preprocessing/                                      # Production preprocessing scripts
├── Frontend/
│   ├── Templates/
│   │   └── index.html                                      # Main application template
│   └── static/
│       ├── css/
│       │   └── styles.css                                 # Custom styling
│       └── js/
│           └── script.js                                  # Frontend logic
├── tests/                                                 # Unit and integration tests
├── Dockerfile                                             # Container configuration
└── requirements.txt                                       # Python dependencies

📚 API Documentation

Translation Endpoint

POST /translate
Content-Type: application/json

Request Body

{
    "source_text": "Hi friends, how are you",
    "source_lang": "en",
    "target_lang": "hi"
}

Response

{
    "translated_text": "नमस्कार दोस्तों, आप कैसे हैं?"
}

Example: Greek Translation

{
    "source_text": "Hello, how are you?",
    "source_lang": "en",
    "target_lang": "el"
}

Response

{
    "translated_text": "Γειά σου, πώς είσαι;"
}

Supported Languages

• English (en)
• Hindi (hi)
• Greek (el)

Error Responses

{
    "error": "Model or tokenizer not loaded."
}

Status: 500

{
    "error": "Translation failed."
}

Status: 500

🤝 Contributing

1. Fork the repository
2. Create your feature branch: git checkout -b feature/NewFeature
3. Commit your changes: git commit -m 'Add NewFeature'
4. Push to the branch: git push origin feature/NewFeature
5. Submit a pull request

📝 License

This project is licensed under the MIT License - see the LICENSE file for details.

---