Toxicity classifier demo showing different result with figure in readme.

[happyhj]

TensorFlow.js version

1.0.0

Browser version

Version 72.0.3626.121 (Official Build) (64-bit)

Describe the problem

Toxicity classifier demo showing different result with figure in readme.
Below the image is from the readme.

In the image, input text "now join the anti gay hitler rebellion now!" results toxicity and identity flag positive.

https://storage.googleapis.com/tfjs-models/demos/toxicity/index.html
But on the demo page, typing input text with "now join the anti gay hitler rebellion now!" makes every flag "negative".

I think the mismatch between this picture and the demo will cause confusion.

Code to reproduce the bug / link to feature request

1. Visit https://storage.googleapis.com/tfjs-models/demos/toxicity/index.html
2. Type "now join the anti gay hitler rebellion now!" on the input area.
3. Click "classify" button.
4. Will show result like following image.