SPEAKER: In this video, you'll see

how you can get TensorBoard working

with Keras-based TensorFlow code.

In this example, I'm using the Fashion MNIST dataset

to do some basic computer vision, where

I can train a Keras neural network to classify

items of clothing.

It's set to train for five epochs,

and you can see the progress, including

the loss and the accuracy, in the output window.

We can see that it finishes training

with an accuracy of about 86%, and we

output some sample predictions.

But how do we visualize this with TensorBoard?

Let's start by importing the time library

and TensorBoard itself.

It can be found in tensorflow.pytho

n.kera.callbacks.

Next, after the model is defined,

we want to instantiate TensorBoard.

Note that we specify a log directory

where stuff will get written.

Finally, as the model is training

in the model.fit function, we need

to tell Keras to call back to TensorBoard.

We simply do this by specifying the callback's parameter

and tell it to use whatever we call the TensorBoard instance.

In this case, it's all lowercase tensorboard.

Now, in your terminal, you can execute the TensorBoard

command, pointing at the log directory

that you just specified.

You'll see that it executes, and it gives me

a TensorBoard at HTTP Machine Name colon 6,006.

Now, if I retrain again, when it's done,

I can take a look in TensorBoard.

TensorBoard will launch, and I can start investigating things

like the loss and the accuracy.

I can also look at the graph that

was built for the training.

And that's just how to get it up and running.

There's lots of great things that you can do further with

TensorBoard, and you can see them

at tensorflow.org/tensorboard.

To learn more about TensorFlow, visit tensorflow.org.

For more videos about TensorFlow,

click the Subscribe button, and if you've

any questions about this video, please leave them

in the comments below.

Thank you.

[MUSIC PLAYING]