Search⌘ K

TensorBoard

Explore how to use TensorBoard to monitor and visualize the training process of TensorFlow models. Learn to track metrics like accuracy and visualize data distributions to ensure model correctness and performance during training.

Chapter Goals:

  • Learn about TensorBoard and how to track the progression of training values
  • Specify training values to be shown in TensorBoard

A. Training visualizations

When training a complex neural network, it is useful to have visualizations of the compuation graph and important values to make sure everything is correct. In TensorFlow, there is a tool known as TensorBoard, which lets us visualize all the important aspects of a model.

Computation graph structure of a neural network shown in TensorBoard.
Computation graph structure of a neural network shown in TensorBoard.

TensorBoard works by reading in an events file, which contains all the model data we want visualized. When training a model, the events file is stored in the same directory as the ...