📝 T5 CNN/DailyMail Summarization

Fine-tuning the T5-small model on the CNN/DailyMail dataset for text summarization using Hugging Face Transformers.
Automatically generate concise news summaries with state-of-the-art Transformers ⚡

---

🚀 Features

• ✅ Preprocessing and tokenization of the dataset
• ✅ Fine-tuning T5-small using Seq2SeqTrainer
• ✅ Evaluation with ROUGE metrics
• ✅ Inference function to generate summaries from any text
• ✅ GPU-friendly (supports mixed precision for faster training)

---

🗂 Project Structure

t5-cnn-dailymail-summarization/
│
├── .gitignore
├── requirements.txt
├── README.md
└── src/
├── train.py       # Sets up Trainer and trains the model
├── inference.py   # Generates summaries / inference
└── config.py      # Stores training\_args and fixed parameters

---

⚙️ Installation 🛠️

Follow these steps to set up the project:

# 1️⃣ Clone the repository
git clone https://github.com/<YOUR_USERNAME>/t5-cnn-dailymail-summarization.git
cd t5-cnn-dailymail-summarization

# 2️⃣ Install dependencies
pip install -r requirements.txt

🏋️ Training

• train.py fine-tunes the T5 model and saves outputs to OUTPUT_DIR.
• All parameters and paths are configurable in src/config.py.

python src/train.py

---

📊 Evaluation

• During training, the model is evaluated using ROUGE scores.
• ROUGE-1, ROUGE-2, and ROUGE-L metrics are computed and displayed automatically.

---

🤖 Inference

• Generate summaries using the trained model:

python src/inference.py

• Example usage inside inference.py:

from src.inference import generate_summary

text = "The Apollo program was the third United States human spaceflight program..."
summary = generate_summary(text)
print(summary)

---

📌 Requirements

• Python >= 3.9
• PyTorch >= 2.0
• transformers >= 4.30
• datasets >= 2.13
• evaluate >= 0.4
• numpy, pandas, scikit-learn, tqdm, accelerate

---

🖼️ Demo / Screenshot

Here is an example of T5 generating a concise summary from a news article:

---