metadata
language: en
license: apache-2.0
base_model: distilbert/distilbert-base-uncased
tags:
- text-classification
- sentiment-analysis
- distilbert
datasets:
- imdb
metrics:
- loss
DistilBERT Sentiment Classifier (IMDB)
Fine-tuned distilbert-base-uncased for binary sentiment classification (positive/negative) on a subset of the IMDB movie review dataset.
Model Details
- Base model: distilbert/distilbert-base-uncased
- Task: Sentiment analysis (binary classification)
- Labels:
0= negative,1= positive - Max sequence length: 128 tokens
Training
| Hyperparameter | Value |
|---|---|
| Dataset | IMDB (500 samples, 80/20 split) |
| Epochs | 2 |
| Batch size | 8 |
| Learning rate | 5e-5 (linear decay) |
Final eval loss: 0.0008
Usage
from transformers import pipeline
classifier = pipeline("text-classification", model="chinmaygarde/hello")
classifier("This movie was absolutely fantastic!")
# [{'label': 'LABEL_1', 'score': 0.999}]