This project enables you to deploy openAI's GPT-2 to production. Using python, flask, Apache server, Docker and Kubernetes.
These instructions will get you a copy of the project up and running on your local machine for development and testing purposes. See deployment for notes on how to deploy the project on a production system.
clone the repo.
git clone https://github.com/emmanuelraj7/opengpt2.git
What things you need to install the software and how to install them
pip3 install -r requirements.txt
and tensorflow-gpu to run in a GPU
pip3 install tensorflow-gpu==1.12.0
After installing needed dependencies
Navigate to /flask_demo
Download model data: downloads models with 124, 355, 774 Million and 1.5 Billion parameters
python3 download_model.py 124M
python3 download_model.py 355M
python3 download_model.py 774M
python3 download_model.py 1558M
And run the development server
python3 flask_predict_api.py
server will run on http://0.0.0.0:5000/ access api documentation - http://0.0.0.0:5000/apidocs
- Containerize the application using docker. Create a container,
docker build --tag openai-gpt2 .
- Push the docker container image to docker hub/registry of your choice from where you can deploy to your target kubernetes cluster or needed deployment target.
docker push "container-id/tag"
To run it locally:
docker run -d -p 8000:8000 containerid
Bind port 8000 of the container to your local machine, as the Apache server in the container is running at port 8000.
- Python - Programming language used
- Flask - The web framework used
- Apache - Server to host the web application
- Docker - Used for containerizing the application
- Kubernetes - Container-orchestration system for deployments, scaling and managment
- Emmanuel Raj - Machine Learning Engineer, Tieto -
This project has MIT license, check LICENSE file for more info and feel free to use it.
- OpenAI team - For releasing the GPT-2 models and code.
- Code for OpenAI's GPT-2 was taken from OpenAI repo and using this I build Flask, Apache and Docker image for deployment.