# RoBERTa-Base Quantized Model for Intent Classification for Banking Systems

This repository contains a fine-tuned RoBERTa-Base model for **intent classification** on the **Banking77** dataset. The model identifies user intent from natural language queries in the context of banking services.

## Model Details

- **Model Architecture:** RoBERTa Base  
- **Task:** Intent Classification  
- **Dataset:** Banking77  
- **Use Case:** Detecting user intents in banking conversations  
- **Fine-tuning Framework:** Hugging Face Transformers  

## Usage

### Installation

```bash
pip install transformers torch datasets
```

### Loading the Model

```python
from transformers import RobertaTokenizerFast, RobertaForSequenceClassification
import torch
from datasets import load_dataset

# Load tokenizer and model
tokenizer = RobertaTokenizerFast.from_pretrained("roberta-base")
model = RobertaForSequenceClassification.from_pretrained("path_to_your_fine_tuned_model")
model.eval()

# Sample input
text = "I am still waiting on my card?"

# Tokenize and predict
inputs = tokenizer(text, return_tensors="pt", truncation=True, padding=True)
with torch.no_grad():
    outputs = model(**inputs)
    predicted_class = torch.argmax(outputs.logits, dim=1).item()

# Load label mapping from dataset
label_map = load_dataset("PolyAI/banking77")["train"].features["label"].int2str
predicted_label = label_map(predicted_class)

print(f"Predicted Intent: {predicted_label}")
```

## Performance Metrics

- **Accuracy:** 0.927922  
- **Precision:** 0.931764  
- **Recall:**  0.927922  
- **F1 Score:** 0.927976  

## Fine-Tuning Details

### Dataset

The Banking77 dataset contains 13,083 labeled queries across 77 banking-related intents, including tasks like checking balances, transferring money, and reporting fraud.

### Training Configuration

- Number of epochs: 5  
- Batch size: 16  
- Evaluation strategy: epoch  
- Learning rate: 2e-5  

## Repository Structure

```
.
├── config.json
├── tokenizer_config.json    
├── special_tokens_map.json 
├── tokenizer.json        
├── model.safetensors       # Fine-tuned RoBERTa model
├── README.md               # Documentation
```

## Limitations
 
- The model may not generalize well to domains outside the fine-tuning dataset.  

- Quantization may result in minor accuracy degradation compared to full-precision models.  
 
## Contributing
 
Contributions are welcome! Feel free to open an issue or submit a pull request if you have suggestions or improvements.