View source on GitHub
|
Core wrapper.
This file contains the Keras model wrapper around an Yggdrasil model/learner. While it can be used directly, the helper functions in keras.py / wrapper_pre_generated.py should be preferred as they explicit more directly the learner specific hyper-parameters.
Usage example:
# Indirect usage
import tensorflow_decision_forests as tfdf
model = tfdf.keras.RandomForestModel()
train_ds = tfdf.keras.pd_dataframe_to_tf_dataset(...)
model.fit(train_ds)
# Direct usage
import tensorflow_decision_forests as tfdf
model = tfdf.keras.CoreModel(learner="RANDOM_FOREST")
train_ds = tfdf.keras.pd_dataframe_to_tf_dataset(...)
model.fit(train_ds)
See "CoreModel" for more details
Classes
class AdvancedArguments: Advanced control of the model that most users won't need to use.
class CoreModel: Keras Model V2 wrapper around an Yggdrasil Learner and Model.
class FeatureSemantic: Semantic (e.g.
class FeatureUsage: Semantic and hyper-parameters for a single feature.
class HyperParameterTemplate: Named and versionned set of hyper-parameters.
class InferenceCoreModel: Keras Model V2 wrapper around an Yggdrasil Model.
class Monotonic: Monotonic constraint between a feature and the model output.
class MultiTaskItem: A single task in a multi-task configuration.
class NodeFormat: Node format of a model.
class YggdrasilDeploymentConfig: A ProtocolMessage
class YggdrasilTrainingConfig: A ProtocolMessage
class datetime: datetime(year, month, day[, hour[, minute[, second[, microsecond[,tzinfo]]]]])
Functions
get_worker_idx_and_num_workers(...): Gets the current worker index and the total number of workers.
no_automatic_dependency_tracking(...): Disables automatic dependency tracking on attribute assignment.
pd_dataframe_to_tf_dataset(...): Converts a Panda Dataframe into a TF Dataset compatible with Keras.
yggdrasil_model_to_keras_model(...): Converts an Yggdrasil model into a TensorFlow SavedModel / Keras model.