Rupa421/BERT-Fine-Tuning

BERT Fine-Tuning for IMDb Sentiment Classification

A fine-tuned BERT Base Uncased model for binary sentiment classification on the IMDb Movie Reviews dataset.

This project demonstrates the complete fine-tuning workflow using the Hugging Face ecosystem, from dataset preprocessing and tokenization to model training, evaluation, inference, and deployment.

---

Model Details

• Base Model: bert-base-uncased
• Task: Binary Sentiment Classification
• Dataset: IMDb Movie Reviews
• Framework: Hugging Face Transformers
• Training Framework: Trainer API
• Language: English

---

Training Pipeline

The model was trained using the following workflow:

• Dataset loading using Hugging Face Datasets
• Tokenization with AutoTokenizer
• Fine-tuning using AutoModelForSequenceClassification
• Evaluation with Accuracy metric
• Mixed precision (FP16) training when CUDA is available
• Model exported using SafeTensors

---

Performance

The fine-tuned model learns to classify movie reviews into:

• LABEL_0 → Negative
• LABEL_1 → Positive

---

Usage

from transformers import pipeline

classifier = pipeline(
    "text-classification",
    model="YOUR_USERNAME/BERT-Fine-Tuning"
)

classifier("This movie was absolutely amazing!")

Example output:

[
    {
        "label": "LABEL_1",
        "score": 0.998
    }
]

---

Repository Contents

• Fine-tuned model weights
• Tokenizer files
• Configuration files
• SafeTensors checkpoint

The complete training notebook, source code, and documentation are available in the accompanying GitHub repository.

---

Future Improvements

• LoRA / PEFT fine-tuning
• Multi-class sentiment classification
• Hyperparameter optimization
• Model quantization
• ONNX and TensorRT deployment
• Production inference benchmarking

---

License

This project is released for educational and research purposes.

---

Built with ❤️ by the author.