Efficient data serving
Well-defined input pipelines can not only greatly reduce the time needed to train models, but also help to better preprocess the training samples to guide the networks toward more performant configurations. In this section, we will demonstrate how to build such optimized pipelines, diving into the TensorFlow tf.data API.
Introducing the TensorFlow Data API
While tf.data has already appeared multiple times in the Jupyter notebooks, we have yet to properly introduce this API and its multiple facets.
Intuition behind the TensorFlow Data API
Before detailing tf.data, we will provide some context to justify its relevance to the training of deep learning models.
Feeding fast and data-hungry models
Neural networks (NNs) are data-hungry methods. The larger the datasets they can iterate on during training, the more accurate and robust these neural networks will become. As we have already noticed in our experiments, training a network is thus a heavy task, which can take hours if...
