Hello at first I would like to thanks for this fantastic courses and offer.
I am still having trouble wrapping my head about the two difference batching and input layer techniques used for MNIST and later in audiobook example . I have slight idea that the reason is the different nature of dataset of two exercises but still can you explain in more detail so i could differentiate and understand the need of using different procedures ?
In the MNIST exercises, we are using the dataset format provided by TensorFlow (through tensorflow-datasets). Whenever you are using one of their datasets, they will be preprocessed in a way similar to our MNIST example: https://www.tensorflow.org/datasets
Whenever you’ve got your data on your computer, you’d approach it as we did in the Business Case example!
The 365 Team