Duplicate Sentence Detection with ALBERT-base-v2

📌 Overview

This repository hosts the quantized version of the ALBERT-base-v2 model for Duplicate Sentence Detection. The model is designed to determine whether two sentences convey the same meaning. If they are similar, the model outputs "duplicate" with a confidence score; otherwise, it outputs "not duplicate" with a confidence score. The model has been optimized for efficient deployment while maintaining reasonable accuracy, making it suitable for real-time applications.

🏗 Model Details

Model Architecture: ALBERT-base-v2
Task: Duplicate Sentence Detection
Dataset: Hugging Face's quora-question-pairs
Quantization: Float16 (FP16) for optimized inference
Fine-tuning Framework: Hugging Face Transformers

🚀 Usage

Installation

pip install transformers torch

Loading the Model

from transformers import AlbertTokenizer, AlbertForSequenceClassification
import torch

device = "cuda" if torch.cuda.is_available() else "cpu"

model_name = "AventIQ-AI/albert-duplicate-sentence-detection"
model = AlbertForSequenceClassification.from_pretrained(model_name).to(device)
tokenizer = AlbertTokenizer.from_pretrained(model_name)

Paraphrase Detection Inference

def predict_duplicate(question1, question2, model):
    inputs = tokenizer(question1, question2, truncation=True, padding="max_length", max_length=128, return_tensors="pt")
    
    # ✅ Move inputs to the same device as the model
    inputs = {key: value.to(device) for key, value in inputs.items()}
    
    with torch.no_grad():  # Disable gradient calculation
        outputs = model(**inputs)
        logits = outputs.logits
    
    # ✅ Get prediction
    probs = torch.softmax(logits, dim=1)
    prediction = torch.argmax(probs, dim=1).item()
 
    # ✅ Output the results
    label_map = {0: "Not Duplicate", 1: "Duplicate"}
    print(f"Q1: {question1}")
    print(f"Q2: {question2}")
    print(f"Prediction: {label_map[prediction]} (Confidence: {probs.max().item():.4f})\n")

# 🔍 Test Example
test_samples = [
    ("How can I learn Python quickly?", "What is the fastest way to learn Python?"),  # Duplicate
    ("What is the capital of India?", "Where is New Delhi located?"),  # Duplicate
    ("How to lose weight fast?", "What is the best programming language to learn?"),  # Not Duplicate
    ("Who is the CEO of Tesla?", "What is the net worth of Elon Musk?"),  # Not Duplicate
    ("What is machine learning?", "How does AI work?"),  # Duplicate
]
for q1, q2 in test_samples:
    predict_duplicate(q1, q2, model)

📊 Quantized Model Evaluation Results

🔥 Evaluation Metrics 🔥

✅ Accuracy: 0.7215
✅ Precision: 0.6497
✅ Recall: 0.5440
✅ F1-score: 0.5922

⚡ Quantization Details

Post-training quantization was applied using PyTorch's built-in quantization framework. The model was quantized to Float16 (FP16) to reduce model size and improve inference efficiency while balancing accuracy.

📂 Repository Structure

.
├── model/               # Contains the quantized model files
├── tokenizer_config/    # Tokenizer configuration and vocabulary files
├── model.safetensors/   # Quantized Model
├── README.md            # Model documentation

⚠️ Limitations

The model may struggle with highly nuanced paraphrases.
Quantization may lead to slight degradation in accuracy compared to full-precision models.
Performance may vary across different domains and sentence structures.

🤝 Contributing

Contributions are welcome! Feel free to open an issue or submit a pull request if you have suggestions or improvements.