spoiler-alert

The model trained from roberta-base on the imdb-spoiler dataset for classification. The model was trained using AutoModelForSequenceClassification.from_pretrained for 3 epochs with a learning rate of 2e-5 and weight decay of 0.01.

Check it out!

Data

imdb-spoiler is a subset of a large-dataset for classifying whether a movie review is a spoiler or not.

Evaluation using the dataset validation split gives:

• F1 0.773021
• Accuracy 0.783275

Project Organization

├── README.md          <- The top-level README for developers using this project.
│
├── notebooks          
|   ├── EDA and Split.ipynb       <- Exploring and split the original data in a small subset
│   ├── Fine-Tuning.ipynb             <- Fine-tuning Roberta-base for Text-Classification
│
└──  requirements.txt   <- The requirements file for reproducing the analysis environment, e.g.

Run Locally

Clone the project

  git clone https://github.com/bhavya-giri/spoiler-alert

Go to the project directory

  cd spoiler-alert

Install dependencies

  pip install requirements.txt

Start the notebook

  jupyter notebook

or open with Juoyter Lab

 jupyter lab