Learn TensorFlow DATA Pipeline | tf.data | Why optimize?

Why? Because your data do not fit in memory, you need pre-processing and you want to decouple loading and pre-processing of data from the distribution of parallel computation, training your model.

A single host performance guide to achieve maximum performance between your CPU/GPU system for training your NN model.

Link to main literature:

#TensorFlowAPI
#InputPipeline

00:00 TF2 Input Pipeline
00:50 ETL process
01:45 Multiple CPU cores
11:05 Summary