LostTech.TensorFlow : API Documentation

Type LSTMAutoRegressor

Namespace tensorflow_estimator.python.estimator.canned.timeseries.estimators

Parent TimeSeriesRegressor

Interfaces ILSTMAutoRegressor

Methods

Properties

Public instance methods

object build_one_shot_parsing_serving_input_receiver_fn(object filtering_length, object prediction_length, object default_batch_size, DType values_input_dtype, bool truncate_values)

Build an input_receiver_fn for export_savedmodel accepting tf.Examples.

Only compatible with `OneShotPredictionHead` (see `head`).
Parameters
object filtering_length
The number of time steps used as input to the model, for which values are provided. If more than `filtering_length` values are provided (via `truncate_values`), only the first `filtering_length` values are used.
object prediction_length
The number of time steps requested as predictions from the model. Times and all exogenous features must be provided for these steps.
object default_batch_size
If specified, must be a scalar integer. Sets the batch size in the static shape information of all feature Tensors, which means only this batch size will be accepted by the exported model. If None (default), static shape information for batch sizes is omitted.
DType values_input_dtype
An optional dtype specification for values in the tf.Example protos (either float32 or int64, since these are the numeric types supported by tf.Example). After parsing, values are cast to the model's dtype (float32 or float64).
bool truncate_values
If True, expects `filtering_length + prediction_length` values to be provided, but only uses the first `filtering_length`. If False (default), exactly `filtering_length` values must be provided.
Returns
object
An input_receiver_fn which may be passed to the Estimator's export_savedmodel.

Expects features contained in a vector of serialized tf.Examples with shape [batch size] (dtype tf.string), each tf.Example containing features with the following shapes:

object build_one_shot_parsing_serving_input_receiver_fn_dyn(object filtering_length, object prediction_length, object default_batch_size, object values_input_dtype, ImplicitContainer<T> truncate_values)

Build an input_receiver_fn for export_savedmodel accepting tf.Examples.

Only compatible with `OneShotPredictionHead` (see `head`).
Parameters
object filtering_length
The number of time steps used as input to the model, for which values are provided. If more than `filtering_length` values are provided (via `truncate_values`), only the first `filtering_length` values are used.
object prediction_length
The number of time steps requested as predictions from the model. Times and all exogenous features must be provided for these steps.
object default_batch_size
If specified, must be a scalar integer. Sets the batch size in the static shape information of all feature Tensors, which means only this batch size will be accepted by the exported model. If None (default), static shape information for batch sizes is omitted.
object values_input_dtype
An optional dtype specification for values in the tf.Example protos (either float32 or int64, since these are the numeric types supported by tf.Example). After parsing, values are cast to the model's dtype (float32 or float64).
ImplicitContainer<T> truncate_values
If True, expects `filtering_length + prediction_length` values to be provided, but only uses the first `filtering_length`. If False (default), exactly `filtering_length` values must be provided.
Returns
object
An input_receiver_fn which may be passed to the Estimator's export_savedmodel.

Expects features contained in a vector of serialized tf.Examples with shape [batch size] (dtype tf.string), each tf.Example containing features with the following shapes:

object build_raw_serving_input_receiver_fn(object default_batch_size, object default_series_length)

Build an input_receiver_fn for export_savedmodel which accepts arrays.

Automatically creates placeholders for exogenous `FeatureColumn`s passed to the model.
Parameters
object default_batch_size
If specified, must be a scalar integer. Sets the batch size in the static shape information of all feature Tensors, which means only this batch size will be accepted by the exported model. If None (default), static shape information for batch sizes is omitted.
object default_series_length
If specified, must be a scalar integer. Sets the series length in the static shape information of all feature Tensors, which means only this series length will be accepted by the exported model. If None (default), static shape information for series length is omitted.
Returns
object
An input_receiver_fn which may be passed to the Estimator's export_savedmodel.

object build_raw_serving_input_receiver_fn_dyn(object default_batch_size, object default_series_length)

Build an input_receiver_fn for export_savedmodel which accepts arrays.

Automatically creates placeholders for exogenous `FeatureColumn`s passed to the model.
Parameters
object default_batch_size
If specified, must be a scalar integer. Sets the batch size in the static shape information of all feature Tensors, which means only this batch size will be accepted by the exported model. If None (default), static shape information for batch sizes is omitted.
object default_series_length
If specified, must be a scalar integer. Sets the series length in the static shape information of all feature Tensors, which means only this series length will be accepted by the exported model. If None (default), static shape information for series length is omitted.
Returns
object
An input_receiver_fn which may be passed to the Estimator's export_savedmodel.

Public static methods

LSTMAutoRegressor NewDyn(object periodicities, object input_window_size, object output_window_size, object model_dir, ImplicitContainer<T> num_features, object extra_feature_columns, ImplicitContainer<T> num_timesteps, ImplicitContainer<T> loss, ImplicitContainer<T> num_units, ImplicitContainer<T> optimizer, object config)

Public properties

object config get;

object config_dyn get;

object model_dir get;

object model_dir_dyn get;

object model_fn get;

object model_fn_dyn get;

object params get;

object params_dyn get;

object PythonObject get;