Compare models for sentiment classification problem
The repo aims to quickly compare several models trained to solve the sentiment classification problem. aclImdb dataset was used for sentiment classification data.
Run
Nvidia docker was used to run notebooks and train models.
To build image run:
docker build -t sentiment_classification .
To start notebook server run:
docker run --volume "repo/path:/code" --gpus all --net host -p 8888:8888 -it --rm sentiment_classification jupyter notebook --allow-root --ip 0.0.0.0 --no-browser
and open browser with http://localhost:8888 url.
There are 2 notebooks under notebooks/ dir:
1-Dataset.ipynb- prepares data,2-Model.ipynb- compares models.
Open above notebooks for further information.