imdb-distilbert-demo

IMDb DistilBERT (Demo)

This model is a fine-tuned version of DistilBERT-base-uncased on the IMDb movie reviews dataset.
It is trained for binary sentiment classification: predicting whether a review is positive or negative.

---

Model Details

• Model type: DistilBERT (transformer encoder)
• Base model: distilbert-base-uncased
• Task: Sentiment analysis (binary classification)
• Language: English
• Fine-tuning dataset: IMDb (25k train / 25k test movie reviews)
• Framework: 🤗 Transformers + Datasets + Trainer API

---

Training Procedure

• Epochs: 1 (demo run; can be extended)
• Batch size: 8 (CPU-friendly, adjust for GPU)
• Learning rate: 2e-5
• Weight decay: 0.01
• Optimizer: AdamW
• Evaluation strategy: Per epoch
• Metrics: Accuracy, Weighted F1

Example results (1 epoch, CPU training):

• Accuracy: ~85% (on test subset of 5k samples)
• Weighted F1: ~85%

⚠️ These are approximate demo numbers. With more epochs + GPU training, performance can reach ~90%+.

---

Usage

You can load the model directly with the Hugging Face pipeline:

from transformers import pipeline

clf = pipeline("text-classification", model="vaibhav2507/imdb-distilbert-demo")

print(clf("This movie was fantastic and the acting was brilliant."))
# [{'label': 'POSITIVE', 'score': 0.99}]

print(clf("Plot holes everywhere. I want my two hours back."))
# [{'label': 'NEGATIVE', 'score': 0.98}]