| license: mit | |
| language: | |
| - en | |
| tags: | |
| - sentiment-analysis | |
| - imdb | |
| - roberta | |
| - text-classification | |
| pipeline_tag: text-classification | |
| base_model: textattack/roberta-base-imdb | |
| # IMDB Sentiment RoBERTa | |
| This repository contains a high-accuracy IMDB sentiment classifier for the 2026 machine learning course task. | |
| The model is based on `textattack/roberta-base-imdb`, a RoBERTa sequence-classification model fine-tuned for IMDB sentiment analysis. | |
| ## Evaluation | |
| - Dataset: `imdb_top_500.csv` | |
| - Accuracy: 98.40% | |
| - Correct: 492 / 500 | |
| - Required minimum accuracy: 0.92 | |
| - Labels: `0 = negative`, `1 = positive` | |
| ## Usage | |
| ```python | |
| from transformers import pipeline | |
| classifier = pipeline( | |
| "sentiment-analysis", | |
| model="ceilf6/imdb-sentiment-roberta", | |
| tokenizer="ceilf6/imdb-sentiment-roberta", | |
| ) | |
| print(classifier("This movie is great and deeply moving.")) | |
| ``` | |
| ## CI/CD | |
| GitHub Actions evaluates the model and uploads this repository only when accuracy is at least `0.92`. | |