This project implements a text summarization model using the Pegasus model from Hugging Face's Transformers library. The model is trained on a summarization dataset and evaluated using the ROUGE metric.

To get started, you'll need to install the required libraries. You can do this using pip:

pip install transformers
pip install transformers[sentencepiece] datasets sacrebleu rouge_score py7zr -q
pip install accelerate>=0.21.0
pip install accelerate -U
pip install transformers[torch]

• Check if a GPU is available.
• Initialize the device (cuda for GPU or cpu).

• Use the Pegasus model (google/pegasus-cnn_dailymail) from Hugging Face.

• Load the dataset for summarization (lighteval/legal_summarization).
• Convert the examples to features suitable for training.

• Set up the training arguments.
• Use the Trainer class to train the model on the dataset.

• Evaluate the model using ROUGE metrics.
• Print the summary of sample dialogues.

• Save the trained model and tokenizer for future use.

• Imports and Setup: Install and import necessary libraries.
• Data Loading: Load and prepare the dataset.
• Model Initialization: Initialize the Pegasus model and tokenizer.
• Training: Set up training arguments and train the model.
• Evaluation: Evaluate the model's performance using ROUGE scores.
• Saving: Save the trained model and tokenizer.

• Ensure you have a GPU available for training as it significantly speeds up the process.
• The provided code is designed to run on Google Colab. Adjustments may be needed for other environments.