get_tf_dataset
- NNDataset.get_tf_dataset(split='train', output_channels=None, is_conditional=False, repeat_y=False, add_c_to_y=False, shuffled=False)[source]
tf.data.Datasetof the desired split.- Parameters
split (
str) – One of train, val, test.output_channels (
Optional[Iterable[str]]) – Channels that should be predicted by the neural network. Defaults to all input channels.is_conditional (
bool) – Whether to add condition information to xrepeat_y (
Union[bool,int]) – Match output length to number of losses (otherwise keras will not work, even if its losses that do not need y).add_c_to_y (
bool) – Append condition to y. Needed for adversarial loss.shuffled (
bool) – Shuffle indices before generating data. Will produce same order every time.
- Returns
The dataset.
- Return type