subaligner.network module

class subaligner.network.Network(secret: Optional[object], input_shape: Tuple, hyperparameters: subaligner.hyperparameters.Hyperparameters, model_path: Optional[str] = None, backend: str = 'tensorflow')

Bases: object

Network factory creates DNNs. Not thread safe since the session of keras_backend is global. Only factory methods are allowed when generating DNN objects.

BI_LSTM = 'bi_lstm'

CONV_1D = 'conv_1d'

LSTM = 'lstm'

TYPES = ['lstm', 'bi_lstm', 'conv_1d']

fit_and_get_history(train_data: numpy.ndarray, labels: numpy.ndarray, model_filepath: str, weights_filepath: str, logs_dir: str, training_log: str, resume: bool) → Tuple[List[float], List[float]]

Fit the training data to the network and save the network model as a HDF file.

Parameters

• {numpy.array} -- The Numpy array of training data. (train_data) –

• {numpy.array} -- The Numpy array of training labels. (labels) –

• {string} -- The model file path. (model_filepath) –

• {string} -- The weights file path. (weights_filepath) –

• {string} -- The TensorBoard log file directory. (logs_dir) –

• {string} -- The path to the log file of epoch results. (training_log) –

• {bool} -- True to continue with previous training result or False to start a new one (default (resume) – {False}).

Returns

tuple – A tuple contains validation losses and validation accuracies.

fit_with_generator(train_data_raw: numpy.ndarray, labels_raw: numpy.ndarray, model_filepath: str, weights_filepath: str, logs_dir: str, training_log: str, resume: bool) → Tuple[List[float], List[float]]

Fit the training data to the network and save the network model as a HDF file.

Parameters

• {list} -- The HDF5 raw training data. (train_data_raw) –

• {list} -- The HDF5 raw training labels. (labels_raw) –

• {string} -- The model file path. (model_filepath) –

• {string} -- The weights file path. (weights_filepath) –

• {string} -- The TensorBoard log file directory. (logs_dir) –

• {string} -- The path to the log file of epoch results. (training_log) –

• {bool} -- True to continue with previous training result or False to start a new one (default (resume) – {False}).

Returns

tuple – A tuple contains validation losses and validation accuracies.

classmethod get_from_model(model_path: str, hyperparameters: subaligner.hyperparameters.Hyperparameters) → subaligner.network.Network

Load model into a network object.

Parameters

• {string} -- The path to the model file. (model_path) –

• {Hyperparameters} -- A configuration for hyperparameters used for training. (hyperparameters) –

classmethod get_network(input_shape: Tuple, hyperparameters: subaligner.hyperparameters.Hyperparameters) → subaligner.network.Network

Factory method for creating a network.

Parameters

• {tuple} -- A shape tuple (input_shape) –

• {Hyperparameters} -- A configuration for hyperparameters used for training. (hyperparameters) –

Returns

Network – A constructed network object.

get_predictions(input_data: numpy.ndarray, weights_filepath: str) → numpy.ndarray

Get a Numpy array of predictions.

Parameters

• {numpy.ndarray} -- The input data (input_data) –

• a Numpy array. (as) –

• {string} -- The weights file path. (weights_filepath) –

Returns

numpy.ndarray – The Numpy array of predictions.

property input_shape

Get the input shape of the network.

Returns

tuple – The input shape of the network.

property layers

Get the layers of the network.

Returns

list – The statck of layers contained by the network

static load_model_and_weights(model_filepath: str, weights_filepath: str, hyperparameters: subaligner.hyperparameters.Hyperparameters) → subaligner.network.Network

Load weights to the Network model.

Parameters

• {string} -- The model file path. (model_filepath) –

• {string} -- The weights file path. (weights_filepath) –

• {Hyperparameters} -- A configuration for hyperparameters used for training. (hyperparameters) –

Returns

Network – Reconstructed network object.

property n_type

Get the type of the network.

Returns

string – The type of the network.

static reset() → None

classmethod save_model_and_weights(model_filepath: str, weights_filepath: str, combined_filepath: str) → None

Combine model and weights and save to a file

Parameters

• {string} -- The path to the model file. (model_filepath) –

• {string} -- The path to the weights file. (weights_filepath) –

classmethod simple_fit(input_shape: Tuple, train_data: numpy.ndarray, labels: numpy.ndarray, hyperparameters: subaligner.hyperparameters.Hyperparameters) → Tuple[List[float], List[float]]

Fit the training data to the network and save the network model as a HDF file.

Parameters

• {tuple} -- A shape tuple (input_shape) –

• {numpy.array} -- The Numpy array of training data. (train_data) –

• {numpy.array} -- The Numpy array of training labels. (labels) –

• {Hyperparameters} -- A configuration for hyperparameters used for training. (hyperparameters) –

Returns

tuple – A tuple contains validation losses and validation accuracies.

classmethod simple_fit_with_generator(input_shape: Tuple, train_data_raw: numpy.ndarray, labels_raw: numpy.ndarray, hyperparameters: subaligner.hyperparameters.Hyperparameters) → Tuple[List[float], List[float]]

Fit the training data to the network and save the network model as a HDF file.

Parameters

• {tuple} -- A shape tuple (input_shape) –

• {list} -- The HDF5 raw training data. (train_data_raw) –

• {list} -- The HDF5 raw training labels. (labels_raw) –

• {Hyperparameters} -- A configuration for hyperparameters used for training. (hyperparameters) –

Returns

tuple – A tuple contains validation losses and validation accuracies.

property summary

Print out the summary of the network.