Multilingual Weighted Normalized Temperature Softmax (WNT-Softmax)

Welcome to the official repository of the Multilingual Weighted Normalized Temperature Softmax (WNT-Softmax), developed by the Computational Intelligence and Biomedicine (ICB) group at the University of Málaga. This project leverages advanced machine learning techniques to enhance language processing tasks across multiple languages.

Environment main settings

• Python 3.10+
• PyTorch 2.0.1

Installation

Clone the repository and install the required packages:

git clone https://github.com/FernandoGD97/WNT-Softmax.git
cd WNT-Softmax

Requirements

pip install -r requirements

Usage

The repository includes scripts for data generation and model training. Here are examples of how to use these scripts:

Data Generation

Generate data for Swedish using a machine translation model (If you don't have direct translation from spanish):

python data_generator.py --lang "sv" --model "Helsinki-NLP/opus-mt-en-sv" --data_path "../data/SympT-EMIST/en/train_df.tsv" --from_spanish false --log_file "../logs/data_generation_sv.log"

Generate data for French:

python data_generator.py --lang "fr" --model "Helsinki-NLP/opus-mt-es-fr" --log_file "../logs/data_generation_fr.log"

Training the WNT-Softmax Model

Train the model with English data for 10 epochs:

python WNT_Softmax_training.py --lang en --epochs 10 --log_file "../logs/wnt-multisapbert.log" --save_model True

Features

• Multilingual Support: Capable of processing and understanding multiple languages.
• Advanced Machine Learning Techniques: Utilizes state-of-the-art models and frameworks.
• High-Performance Computing: Initially deployed and tested on an NVIDIA RTX 2080 Ti GPU, though not a mandatory requirement.

Contributing

We welcome contributions from the community. Please submit your pull requests to the main branch and ensure your code adheres to the existing style.