Model

class mouffet.models.model.Model(opts=None)

Base abstract class for defining models. Must be subclassed.

NAME

Name of the model class. Will be used to identify and save classes

STEP_TRAINING

The name of the training step. Will be used in configuration file

STEP_VALIDATION

The name of the validation step. Will be used in configuration file

abstract create_model()

Abstract method where model creation / network initialization should take place

Raises:

NotImplementedError – [description]

property opts

Property that contains the options related to the model as read in the configuration file

abstract predict(x)

Predict data using the model

Parameters:

• x (data) – the data on which we want to makae a prediction. This function should make only

• needed (one prediction and be called numerous times if) –

Raises:

NotImplementedError – [description]

prepare_data(data)

Prepare the data before training. Allows to perform last minute changes to the data just before training.

Parameters:

data (Object) – The data to be prepared

Returns:

The modified data

Return type:

Object

abstract save_model(path=None)

Save the model to disk

Parameters:

path (Object, optional) – Path where the model should be saved. Defaults to None.

Raises:

NotImplementedError – Should be inherited by the final model

save_options(file_name, options)

Save the options related to the model for logging purposes as a yaml file. By default, uses the “results_save_dir” property from the ModelOptions class associated to the model. By default, this will be the following combination: model_dir/model_id/version where model_dir and model_id can be found in the model configuration file and version is calculated automatically.

Parameters:

• file_name (string or pathlib.Path) – The name of the file to be saved

• options (Object) – The options to save that can be transcribed as a yaml file