sethah/sentiment-explorer

An interactive web app for exploring transfer learning for sentiment analysis

Download models

Download models to host so that they don't need to be downloaded on startup every time.

mkdir -p /$HOME/.models
mkdir -p /$HOME/.pytorch_pretrained_bert
pushd /$HOME/.models
curl -O https://sentiment-explorer.s3-us-west-1.amazonaws.com/nbsvm_imdb_sent_500.pkl
curl -O https://sentiment-explorer.s3-us-west-1.amazonaws.com/bert_base_1000.tar.gz
popd

Running with Docker

Run with docker-compose (webapp + nginx)

docker-compose build
docker-compose up

Run just the web app with docker

cd sentiment
nvidia-docker build -t sentiment-explorer:latest .
nvidia-docker run -p 5000:5000 \
-v $HOME/.models/:/models \
-v $HOME/.pytorch_pretrained_bert/:/root/.pytorch_pretrained_bert \
sentiment-explorer:latest

Run without docker

Create a Python virtual environment (Python 3.6 or later)

cd sentiment
pip install -r requirements.txt
gunicorn --timeout 60 --bind 0.0.0.0:5000 wsgi