SagarFadatare00/distilbert-imdb-sentiment

DistilBERT IMDb Sentiment Analysis

A fine-tuned DistilBERT model for classifying movie reviews as positive or negative. This model was trained on the IMDb movie review dataset using Hugging Face Transformers and PyTorch.

Model Details

• Model architecture: DistilBERT
• Base model: distilbert/distilbert-base-uncased
• Task: Binary sentiment classification
• Dataset: IMDb Movie Reviews
• Framework: PyTorch
• Training approach: Fine-tuning a pretrained model
• Labels: NEGATIVE and POSITIVE

Example Predictions

Input Text	Predicted Sentiment
This movie was absolutely fantastic!	POSITIVE
The acting was terrible and the story was boring.	NEGATIVE
I enjoyed every minute of this film.	POSITIVE

Evaluation Results

The model was fine-tuned for one epoch using 5,000 training examples and evaluated on a selected subset of 1,000 IMDb test examples.

Metric	Result
Accuracy	86.8%
F1 Score	87.0%
Evaluation Loss	0.3307

These results reflect the selected evaluation subset and are not a benchmark for all sentiment-analysis tasks.

How to Use

Install the required libraries:

pip install torch transformers

Load the model directly from Hugging Face:

from transformers import pipeline

model_id = "SagarFadatare00/distilbert-imdb-sentiment"

classifier = pipeline(
    "text-classification",
    model=model_id,
)

texts = [
    "This movie was absolutely fantastic!",
    "The acting was terrible and the story was boring.",
]

for text in texts:
    result = classifier(text)[0]
    print("Text:", text)
    print("Sentiment:", result["label"])
    print("Confidence:", round(result["score"] * 100, 2), "%")

Training Configuration

• Learning rate: 2e-5
• Training batch size: 8
• Evaluation batch size: 8
• Number of epochs: 1
• Weight decay: 0.01
• Maximum token length: 256
• Training examples: 5,000
• Evaluation examples: 1,000

The training examples were selected from the IMDb training split, and the evaluation examples were selected from its test split using a fixed random seed.

Limitations

• The model was trained on movie reviews and may not perform equally well on social media posts, product reviews, or other domains.
• It predicts only positive or negative sentiment; it does not classify neutral sentiment as a separate category.
• Prediction confidence is not a guarantee of correctness.
• Performance may vary depending on the text, writing style, and domain.

Intended Use

This model is intended for educational purposes, experimentation, and sentiment-analysis applications. Evaluate it on your own data before using it in production.

Acknowledgments

• Hugging Face Transformers
• DistilBERT
• IMDb Dataset

Author

Sagar Fadatare