TensorFlow has a powerful built-in visualization tool called TensorBoard. It allows developers to interpret, visualize, and debug computational graphs. To visualize graph and metrics automatically in TensorBoard, TensorFlow writes events related to the execution of a computational graph to a particular folder.
This example shows a computational graph of the analysis done earlier:
To examine the graph, click on the graph tab on the top panel of TensorBoard. If the graph has several nodes, visualizing it in a single view can be hard. To make our visualization more accessible, we can organize the related operations into groups using
tf.name_scope with specific names.