Module: tfm.core

Core is shared by both nlp and vision.


actions module: Provides TFM orbit actions and associated helper functions/classes.

base_task module: Defines the base task abstraction.

base_trainer module: Standard Trainer implementation.

config_definitions module: Common configuration settings.

exp_factory module: Experiment factory methods.

export_base module: Base class for model export.

file_writers module: File writer functions for dataset preparation, infra validation, and unit tests.

input_reader module: A common dataset reader.

registry module: Registry utility.

savedmodel_checkpoint_manager module: Custom checkpoint manager that also exports saved models.

task_factory module: A global factory to register and access all registered tasks.

tf_example_builder module: Builder class for preparing tf.train.Example.

tf_example_feature_key module: Data classes for tf.Example proto feature keys.

train_lib module: TFM common training driver library.

train_utils module: Training utils.