14 TensorBoard: Big brother of TensorFlow

This chapter covers,

Running the TensorBoard, visualize and verify label correctness of images in the TensorBoard
Generating model related visualizations in real time depicting model performance and layer activations
Profiling models to understand where performance bottlenecks are using the TensorBoard
Using tf.summary to log custom metrics during customized model training
Visualizing and analyzing word vectors on TensorBoard

Up until chapter 12 we have been focusing on various models. We have talked about fully-connected models (e.g. autoencoders), convolutional neural networks, recurrent neural networks (e.g. LSTMs, GRUs). In chapter 12, we talked about Transformers; a powerful family of deep learning models that have paved the way to a new state of the art performance in language understanding. Furthermore, inspired by the achievements in the field of natural language processing, Transformers are making waves in the computer vision field. Past the modelling step, we still have to plough through several steps to reap the final harvest. One such step is making sure the data/features to the model are correct and models are working as expected.

14.1 Visualize data with the TensorBoard

14.2 Tracking and Monitoring models with TensorBoard

14.3 Using tf.summary to write custom metrics during model training

14.4 Profiling models to detect performance bottlenecks

14.4.1 Optimizing the input pipeline

14.4.2 Mixed precision training

14.5 Visualizing word vectors with the TensorBoard

14.6 Summary