radi-cho / tfjs-firebase

Train a model with data from Firestore, save it to Cloud Storage and make predictions in Cloud Functions - entirely using NodeJS

Bag of words text classification

This workshop covers:

• Training a model with Tensorflow.js inside Cloud Functions using data from Firestore (JavaScript only).
• Retraining a model; Saving and loading a model.
• Making predictions with a cloud function.

The developed model can predict if a comment is positive or negative. We train it with data from the Firestore database. The code can be easily modified to predict star reviews, app statistics, etc.

Demo

Training

Predictions

tfjs-firebase's Issues

question: how do you configure and determine the layer with dense against the vocab?

model.add(tf.layers.dense({ units: 2, inputShape: [vocabulary.length] }));
model.add(tf.layers.dense({ units: 1, inputShape: [2] }));

if i add new vocab how can i reconfigure the dense layer?

Please add a LICENSE

This repository does not have a LICENSE which means we can no longer list it on https://devlibrary.withgoogle.com/ and makes it harder for people to use with confidence.

If you add a license (Apache 2.0 or MIT) please file an issue on https://github.com/google/devlibrary so that we can add your project to the site.