USAGE.md

How the hell do I run this?

Prerequisites

• Git
• Node - If you use a node version manager like nvm / fnm you should be prompted with the correct version through the provided .node-versions file
• Yarn
• Conda/Miniconda - Miniconda is prefered due to it's small footprint
• An Auth0 account
• Docker Engine
• Docker Compose

A few pointers

• You will need to setup an Auth0 account, create an application and an API
• You will only be able to access the web-client and the api endpoints as an authenticated user
• There are several .env files that you need to add secrets (your own) to, they are ignored by default and shoulder never be committed to your remote repository
  • frontend/.env
  • backend/api-gateway/.env
  • backend/audio-to-midi-worker/.env
  • backend/docker/local/.env.local
  • backend/separation-worker/.env
  • backend/transcription-worker/.env
• Use the .env.template/.env.local.template files as a templates
• The setup-files for conda and pip are currently only tested for CUDA devices, in theory if no GPU is detected the CPU should be used as fallback

Running workers with GPU acceleration (CUDA)

Direct - Install dependencies locally (using conda or pip) and start the workers directly.

Containerized

• NVIDIA GPU Drivers
• NVIDIA Container Toolkit

Running workers with GPU acceleration (MacOs - Metal API)

Direct - Install dependencies locally (using conda or pip) and start the workers directly.

Containerized - TBD - Currently no setup files available to run this application containerized with GPU acceleration on arm macs.

Running workers on CPU

Direct - Install dependencies locally (using conda or pip) and start the workers directly.

Containerized - TBD - Currently no setup files available to run this application CPU only.

Continuing on

Check out the documentation for the individual system components:

• Web-Client
• Backend
  • API-Gateway
  • Separation Worker
  • Audio To Midi Worker
  • Transcription Worker