mlreflect.models package

Submodules

mlreflect.models.model_helpers module

mlreflect.models.model_helpers.create_save_directory(directory_name: str)

mlreflect.models.model_helpers.make_save_path(model_directory: str, time_stamp: str)

mlreflect.models.model_helpers.make_tensorboard_callback(model_directory: str, time_stamp: str, graphs_subdirectory: str = 'tb_graphs')

mlreflect.models.simple_model module

class mlreflect.models.simple_model.SimpleModel(directory_name: str, n_input: int, n_output: int)

Bases: object

train(input_train: numpy.ndarray, output_train: Union[pandas.core.frame.DataFrame, numpy.ndarray], input_val: numpy.ndarray, output_val: Union[pandas.core.frame.DataFrame, numpy.ndarray], epochs=60)

train_with_generator(data_generator_train, data_generator_val, epochs=60)

mlreflect.models.trained_model module

class mlreflect.models.trained_model.DefaultTrainedModel

Bases: mlreflect.models.trained_model.TrainedModel

Populated TrainedModel container for a single layer on a Si/SiOx substrate.

class mlreflect.models.trained_model.TrainedModel

Bases: object

Container object for trained keras models. Contains other information necessary to predict reflectivity data.

Object can be loaded from variables or from a saved .h5 file.

Properties:

sample: keras_model: q_values: ip_mean: ip_std:

from_file(file_name: str, **kwargs)

Populates the TrainedModel container with data saved in the given .h5 file.

Parameters

file_name – Full path to the .h5 file that contains the saved data.

from_variable(model: tensorflow.python.keras.engine.training.Model, sample: mlreflect.data_generation.multilayer.MultilayerStructure, q_values: numpy.ndarray, ip_mean: numpy.ndarray, ip_std: numpy.ndarray)

Populates the TrainedModel container with the given arguments.

Parameters

• model – Trained keras model.

• sample – MultiLayer object of the sample structure that what used for training

• q_values – q values that were used for training

• ip_mean – Mean values of all training inputs used for input standardization.

• ip_std – Standard deviation of all training inputs used for input standardization.

property ip_mean

property ip_std

property keras_model

property q_values

property sample

save_model(file_name: str)

Saves the data in the TrainedModel container to an .h5 file.

Parameters

file_name – Full path to the .h5 file.

Module contents

class mlreflect.models.DefaultTrainedModel

Bases: mlreflect.models.trained_model.TrainedModel

Populated TrainedModel container for a single layer on a Si/SiOx substrate.

class mlreflect.models.TrainedModel

Bases: object

Container object for trained keras models. Contains other information necessary to predict reflectivity data.

Object can be loaded from variables or from a saved .h5 file.

Properties:

sample: keras_model: q_values: ip_mean: ip_std:

from_file(file_name: str, **kwargs)

Populates the TrainedModel container with data saved in the given .h5 file.

Parameters

file_name – Full path to the .h5 file that contains the saved data.

from_variable(model: tensorflow.python.keras.engine.training.Model, sample: mlreflect.data_generation.multilayer.MultilayerStructure, q_values: numpy.ndarray, ip_mean: numpy.ndarray, ip_std: numpy.ndarray)

Populates the TrainedModel container with the given arguments.

Parameters

• model – Trained keras model.

• sample – MultiLayer object of the sample structure that what used for training

• q_values – q values that were used for training

• ip_mean – Mean values of all training inputs used for input standardization.

• ip_std – Standard deviation of all training inputs used for input standardization.

property ip_mean

property ip_std

property keras_model

property q_values

property sample

save_model(file_name: str)

Saves the data in the TrainedModel container to an .h5 file.

Parameters

file_name – Full path to the .h5 file.