Module: tfr.keras.pipeline

Ranking pipeline to train tf.keras.Model in tfr.keras.


class AbstractDatasetBuilder: Interface for datasets and signatures.

class AbstractPipeline: Interface for ranking pipeline to train a tf.keras.Model.

class BaseDatasetBuilder: Builds datasets from feature specs.

class DatasetHparams: Hyperparameters used in BaseDatasetBuilder.

class ModelFitPipeline: Pipeline using to train a ranking tf.keras.Model.

class MultiLabelDatasetBuilder: Builds datasets for multi-task training.

class MultiTaskPipeline: Pipeline for multi-task training.

class NullDatasetBuilder: A no-op wrapper of datasets and signatures.

class PipelineHparams: Hyperparameters used in ModelFitPipeline.

class SimpleDatasetBuilder: Builds datasets from feature specs with a single label spec.

class SimplePipeline: Pipleine for single-task training.