Hands-On Machine Learning with TensorFlow.js

Chrome profiler

Google Chrome also provides a powerful tool that we can use to inspect the performance of our web applications. Although it is designed for general web applications, we can use it to profile our machine learning application. After launching the window for the profiler, click on the Performance tab. This tool calculates the time each function call of the JavaScript runtime takes to complete. The following screenshot shows the call diagram of the benchmark we used previously:

As we can see, tf.matMul took 1.65 ms to complete. By zooming into the image, we can find out what tf.matMul does. As shown in the following screenshot, ...