Tensorflowjs simple LSTM example

Intro: This example aims to test the limitations of "in browser learning". What I learned during this little project is how difference the flow in JavaScript compared to the Python version of Tensorflow. Memory handling is a tricky but neccessary part of the JavaScript version. WebGL has its own limitations. GPU device can not be targeted directly. In dual GPU notebooks (integrated - dedicated) the dedicated GPU should be set explicitly in the card manufacturer settings. It's not magic, the learning just works, but it utilizes limited resources.

https://js.tensorflow.org/

This experiment was based on this article:

https://towardsdatascience.com/lstm-by-example-using-tensorflow-feb0c1968537

For test run

1. Clone repository
2. Open index.html in broswser. The already bundled js will load, so the parameters could not be changed.

For development

• Install local dependencies

  npm install

• Install webpack globally

  npm install webpack webpack-cli -g

• Install http-server for loading assets, etc.

  npm install http-server -g

Finally run:

• npm run webpack
• http-server

in separete terminals.

Experiment with it!