GPU Server WebApp

This repository contains the code for the web application, meaning, the frontend (user visible) and the backend (database etc interaction). It is written in Next.js.

First setup

Before you can develop this repo locally, you follow a few steps:

1. Install PostgreSQL.
2. Copy the .env.example to .env and fill in the properties. (see Discord for credentials)
3. Install NodeJS
4. Run npm install to install the dependencies.

Local development

1. Make sure the PostgreSQL database is running.
2. Run npm run dev, the web application will start. Go to https://localhost:5001 in your browser. If you make a code change, this will immediately be updated in the browser.

Database modifications

When you adjust the database schema (prisma/schema.prisma), you must update the local database using migrations.

Run the following commands to update your local database:

• npx prisma migrate dev
• npx prisma generate

Dockerfile tensorflow/federated

This repo uses the tensorflow/federated repository and its Dockerfile. The build version of this Dockerfile is not officially stored in docker hub. Because of this a custom built version of the Dockerfile is located here.

docker pull codestix/federated

Build and publish new version

1. Create a new repository or use existing repository on hub.docker.com.

2. Login using the docker login command.

3. Clone the tensorflow/federated git repository ...

   git clone https://github.com/tensorflow/federated
   cd federated
   

4. Build the image using ... (make sure to replace versions and tag)

   docker build --build-arg VERSION="0.23.0" --network=host  --file "tensorflow_federated/tools/runtime/container/release.Dockerfile" -t codestix/federated:0.23.0 .
   

5. Push the built image to Docker Hub ... (make sure to replace tag and version)

   docker push codestix/federated:0.23.0