DistilBERT Sentiment Analysis
Fine-tuned distilbert-base-uncased on the IMDB movie review dataset for binary sentiment classification (positive / negative).
Results
| Metric | Score |
|---|---|
| Test accuracy | 91.33% |
| Training loss | 0.238 |
| Validation loss | 0.225 |
| Epochs | 1 |
Usage
from transformers import pipeline
classifier = pipeline(
"text-classification",
model="BlessedDovis/distilbert-sentiment-analysis"
)
result = classifier("This movie was absolutely fantastic!")
# POSITIVE — 99.52% confidence
Training
- Base model: distilbert-base-uncased
- Dataset: IMDB (25,000 train / 25,000 test)
- Hardware: Google Colab T4 GPU
- Framework: HuggingFace Transformers + PyTorch
- Batch size: 16
- Mixed precision: fp16
- Downloads last month
- 1