Help protect the Great Barrier Reef with TensorFlow on Kaggle Join Challenge

tff.learning.reconstruction.simple_dataset_split_fn

An example of a DatasetSplitFn that returns the original client data.

Both the reconstruction data and post-reconstruction data will result from iterating over the same tf.data.Dataset. Note that depending on any preprocessing steps applied to client tf.data.Datasets, this may not produce exactly the same data in the same order for both reconstruction and post-reconstruction. For example, if client_dataset.shuffle(reshuffle_each_iteration=True) was applied, post-reconstruction data will be in a different order than reconstruction data.

client_dataset tf.data.Dataset representing client data.

A tuple of two tf.data.Datasets, the first to be used for reconstruction, the second to be used post-reconstruction.