Loading data into AutoKeras in multiple formats
As we mentioned previously, AutoKeras performs normalization automatically. However, in the following chapters, you will see that you can create your model in a more personalized way by stacking blocks. More specifically, you can use special blocks to normalize your data.
Now, let's look at the different data structures that we can use to feed our model.
AutoKeras models accept three types of input:
- A NumPy array is an array that's commonly used by Scikit-Learn and many other Python-based libraries. This is always the fastest option, as long as your data fits in memory.
- Python generators load batches of data from disk to memory, so this is a good option when the entire dataset does not fit in memory.
- TensorFlow Dataset is a high-performance option that is similar to Python generators, but it is best suited for deep learning and large datasets. This is because data can be streamed from disk or from a distributed...
