Sentiment Analyzer
Overview
This repository contains a sentiment analysis model trained on the IMDb dataset. The model is based on distilbert-base-uncased and fine-tuned for binary sentiment classification (positive/negative).
Files
sentiment_model/: Contains the trained model files.results/checkpoint-125/: Checkpoint directory from training.sample_data/: Sample dataset files used for training and evaluation.wandb/: Weights & Biases logs and run data.
Installation
pip install transformers datasets sentence-transformers
Usage
from transformers import AutoTokenizer, AutoModelForSequenceClassification
tokenizer = AutoTokenizer.from_pretrained('kaniskaZoro/sentiment-analyzer')
model = AutoModelForSequenceClassification.from_pretrained('kaniskaZoro/sentiment-analyzer')
text = "The movie was fantastic!"
inputs = tokenizer(text, return_tensors='pt', padding=True, truncation=True, max_length=256)
outputs = model(**inputs)
Training
The model was trained using the Trainer API from Hugging Face Transformers with the following settings:
- Dataset: IMDb (subset of 2000 train, 500 test samples for demonstration)
- Batch size: 16
- Epochs: 1
- Logging and checkpointing integrated with Weights & Biases
License
Specify license here (e.g., MIT, Apache 2.0).