This is a (work-in-progress) translation of a portion of harmonic distance and
chord rationalization research into TensorFlowJS, from a previous implementation
in Python. This aims to allow the running of the harmonic distance and chord
rationalization algorithms on npm, for use in Max 8 and possibly the browser.
Make sure to set the proper environment variable, given here for PowerShell.
$Env:TF_FORCE_GPU_ALLOW_GROWTH = "true"
Assumes you have a typescript compiler installed, as this will run tsc
npm install -g ts-node typescript
# Install nodejs and npm on whatever's right for your system
sudo apt install nodejs npm
# Install yarn for packaging
sudo npm install -g yarn
# Install all the packages
yarn install
# Build the typescript files to dist/
yarn build
There are lots of instances where you might get an OOM error, so I've included a
stress_test.ts script to help identify the choke-points. If you get an error,
please open an issue (including the params you called new Minimizer with) so
that I can implement batch-processing where possible.
To run with typescript:
node -r ts-node/register ./stress-test.ts
Alternatively, if you're using a debugger like VSCode and want to break on entry:
node -r ts-node/register --inspect --inspect-brk ./stress-test.ts