Intent Detection using XLM-Roberta

Overview

This repository contains code and resources for building an Intent Detection model using the XLM-Roberta architecture. Intent detection is a common task in Natural Language Processing (NLP), where the goal is to determine the intent or purpose behind a user's input.

In this project, we leverage the power of the Hugging Face Transformers library to fine-tune an XLM-Roberta model for intent detection tasks. The model can be used for various applications, including chatbots, virtual assistants, and customer support systems.

Features

• Preprocessing scripts for data preparation.
• Fine-tuning scripts for training the XLM-Roberta model on your dataset.
• Inference code for predicting intents using the trained model.
• Example notebooks and usage instructions.

Getting Started

Prerequisites

Before getting started, ensure you have the following prerequisites:

• Python 3.x
• PyTorch
• Hugging Face Transformers library
• Other dependencies as specified in requirements.txt

Installation

1. Clone this repository to your local machine:

   git clone https://github.com/MLArtist/intent-detection-using-XLM-Roberta.git
   cd intent-detection-using-XLM-Roberta

2. Install the required dependencies:

   pip install -r requirements.txt

Usage

Option 1: Bare Metal

To perform inference on your local machine without using Docker, follow these steps:

1. Download the fine-tuned model weights from the Google Drive.

2. Place the downloaded files in the 'xlm-roberta-large-custom-trained' folder.

3. Once you have completed these steps, run the inference server:

   python -m server --model xlm-roberta-large-custom-trained

4. Test the inference server

   curl http://localhost:8080/ready
   curl -X POST -H "Content-Type: application/json" -d '{"text": "when is the next flight to new york"}' http://localhost:8080/intent

Option 2: Via Docker

To perform inference using Docker, follow these steps:

1. Download the fine-tuned model weights from the Google Drive.

2. Place the downloaded files in the 'xlm-roberta-large-custom-trained' folder.

3. Build the Docker image:

   docker build -t my-roberta-app -f Dockerfile .

4. Run the Docker container, mapping port 8080 (or your preferred port) and specifying the model:

   docker-compose -f docker-compose.yml up

5. test the inference server

   curl http://localhost:8080/ready
   curl -X POST -H "Content-Type: application/json" -d '{"text": "when is the next flight to new york"}' http://localhost:8080/intent

Training Instructions

Use the example notebook Train.ipynb in the notebooks/ directory for detailed training and offline model evaluation.

Acknowledgments

• Hugging Face Transformers

Authors

• Amit Kumar

For any inquiries or issues, please open an issue.