NLP with PyTorch: How-to Generator

This project containes a GPT-2 transformer fine-tuned on WikiHow.com entries related to cultural practices. You can test the model, which is currenlty deployed on a DigitalOcean instace here: http://how-to-generator.herokuapp.com/

The ideation and UI has been done in collaboration with Johanna K. Michel

Steps to Deploy Model served via FastAPI using docker

Ref: https://github.com/tiangolo/uvicorn-gunicorn-fastapi-docker

1. Dockerize FastAPI App

• Add Docker file vim Dockerfile and paste:

FROM tiangolo/uvicorn-gunicorn-fastapi:python3.8

COPY ./requirements.txt /app/requirements.txt

RUN pip install --no-cache-dir --upgrade -r /app/requirements.txt

COPY ./app /app

2. Create Docker Image

• Build docker image docker build -t myimage ./

• Run Docker image locally (for testing) docker run -p 4000:80 myimage

3. Push to Docker Hub

• Login to Docker docker login

• Tag image and Push docker tag myimage plasticfruits/reponame:latest docker push plasticfruits/reponame:latest

4. Setup Container and run

For large models minimum 4GB of RAM

• Create RancheraOS and SSH ssh rancher@your_droplet_ip

• pull Docker Image docker pull plasticfruits/how-to-qa

• Check docker image info docker images

• Run Container docker run -d --name mycontainer -p 80:80 plasticfruits/how-to-qa

• Check if container is running docker ps docker ps -a for all containers

• Check logs docker container logs mycontainer