js-training | Get good at JavaScript | Runtime Evironment library

Formik itself has many asynchronous method and hook inside it, therefore in your test you also need to write in an asynchronous way. Speaking of testing hook in general, i recommend you use react-testing-library together with enzyme and jest. You can have a glance here: https://testing-library.com/docs/react-testing-library/intro

For more detail on how to use it specifically with Formik. I recommend you read this article, quite a good explanation. ( Or can google 'test formik with react-testing library')

Briefly, you need to cover it in a wait method of the library. Something like this:

The ts typing of axis indicates that it is a number, thus the complain of tslint. Looking at the source code, the axis is later pushed into an array with the following

The prediction values differs at each refresh because at each refresh there is a new training. The model weights are initialized with random values. During the training, the weights can either converge to optimum values or diverge. It depends on many parameters. Actually even the best model will not always converge at a fixed number of training epochs.

To always have the same value, the initial values of the weights can be set with fixed data. But again how to find this data that will lead to a best prediction ? That's not always easy to find those weights. A simple way to initialize a layer weights would be to use the kernelInitializer of the layer.

The image should be predicted only when it has completed to load