tfr.keras.pipeline.DatasetHparams

Hyperparameters used in BaseDatasetBuilder.

tfr.keras.pipeline.DatasetHparams(
    train_input_pattern: str,
    valid_input_pattern: str,
    train_batch_size: int,
    valid_batch_size: int,
    list_size: Optional[int] = None,
    valid_list_size: Optional[int] = None,
    dataset_reader: Any = tfr.keras.pipeline.DatasetHparams.dataset_reader,
    convert_labels_to_binary: bool = False
)

Hyperparameters to be specified to create the dataset_builder.

train_input_pattern A glob pattern to specify the paths to the input data for training.
valid_input_pattern A glob pattern to specify the paths to the input data for validation.
train_batch_size An integer to specify the batch size of training dataset.
valid_batch_size An integer to specify the batch size of valid dataset.
list_size An integer to specify the list size. When None, data will be padded to the longest list in each batch.
valid_list_size An integer to specify the list size in valid dataset. When not specified, valid dataset uses the same list size as list_size.
dataset_reader A function or class that can be called with a filenames tensor and (optional) reader_args and returns a Dataset. Defaults to tf.data.TFRecordDataset.
convert_labels_to_binary A boolean to indicate whether to use binary label.

Child Classes

class dataset_reader

Methods

__eq__

__eq__(
    other
)

convert_labels_to_binary False
list_size None
valid_list_size None