pywatts.modules.wrappers package

Submodules

pywatts.modules.wrappers.base_wrapper module

class pywatts.modules.wrappers.base_wrapper.BaseWrapper(name: str)

Bases: pywatts.core.base.BaseEstimator, abc.ABC

The base wrappers class, where all wrappers have to inherit from.

Parameters:name (str) – Name of the module

pywatts.modules.wrappers.dl_wrapper module

class pywatts.modules.wrappers.dl_wrapper.DlWrapper(model, name, fit_kwargs=None)

Bases: pywatts.modules.wrappers.base_wrapper.BaseWrapper, abc.ABC

Super class for deep learning framework wrappers

Parameters:
  • model – The deep learning model
  • name (str) – The name of the wrappers
  • fit_kwargs (dict) – The fit keyword arguments necessary for fitting the model

pywatts.modules.wrappers.function_module module

class pywatts.modules.wrappers.function_module.FunctionModule(transform_method: Callable, fit_method: Optional[Callable] = None, name: str = 'FunctionModule')

Bases: pywatts.core.base.BaseEstimator

This module calls the function in its transform and optional fit method. It can be used, for executing own code in the pipeline. Note that the wrapped function is called with the same keyword arguments as the module.

Parameters:
  • transform_method (Callable) – The function which should be executed in the transform step.
  • fit_method (Callable) – The function which should be executed in the fit step.
  • name (str) – name of the instance (FunctionModule)
fit(**kwargs) → xarray.core.dataarray.DataArray

Call the fit_method if available wrapped by this module on x.

Parameters:kwargs (xr.DataArray) – The input arrays
get_params() → Dict[str, object]

Returns an empty dictionary, since this wrappers does not contain any parameters

Returns:Empty dictionary
Return type:dict
classmethod load(load_information)

Loads a condition from a JSON file and creates the corresponding Conditional

Parameters:load_information (Dict) – JSON file of the Conditional
Returns:conditional module from file
Return type:Conditional

Warning

This method use pickle for loading the module. Note that this is not safe. Consequently, load only modules you trust. For more details about pickling see https://docs.python.org/3/library/pickle.html

save(fm: pywatts.core.filemanager.FileManager)

Saves the Conditional module to JSON file

Parameters:fm (FileManager) – A FileManager, from which the path where the JSON file is saved is fetches
Returns:Dictionary with name, parameters, related module and class, and path to the file
Return type:Dict
set_params(*args, **kwargs)

Does nothing:

transform(**kwargs) → xarray.core.dataarray.DataArray

Call the transform_method wrapped by this module on x.

Parameters:x (xarray.DataArray) – The input arrays
Returns:The transformed DataArray
Return type:xarray.DataArray

pywatts.modules.wrappers.keras_wrapper module

class pywatts.modules.wrappers.keras_wrapper.KerasWrapper(model: Union[keras.engine.training.Model, Tuple[keras.engine.training.Model, Dict[str, keras.engine.training.Model]]], name: str = 'KerasWrapper', fit_kwargs=None, compile_kwargs=None, custom_objects=None)

Bases: pywatts.modules.wrappers.dl_wrapper.DlWrapper

Wrapper class for keras models

Parameters:
  • model – The deep learning model
  • name (str) – The name of the wrappers
  • fit_kwargs (dict) – The fit keyword arguments necessary for fitting the model
  • compile_kwargs (dict) – The compile keyword arguments necessary for compiling the model.
  • custom_objects (dict) – This dict contains all custom objects needed by the keras model. Note, users that uses such customs objects (e.g. Custom Loss) need to specify this to enable the loading of the stored Keras model.
fit(**kwargs)

Calls the compile and the fit method of the wrapped keras module. :param x: The input data :param y: The target data

get_params() → Dict[str, object]

Returns the parameters of deep learning frameworks. :return: A dict containing the fit keyword arguments and the compile keyword arguments

classmethod load(load_information) → pywatts.modules.wrappers.keras_wrapper.KerasWrapper

Load the keras model and instantiate a new keraswrapper class containing the model. :param params: The paramters which should be used for restoring the model. (Note: This models should be taken from the pipeline json file) :return: A wrapped keras model.

save(fm: pywatts.core.filemanager.FileManager) → dict

Stores the keras model at the given path :param fm: The Filemanager, which contains the path where the model should be stored :return: The path where the model is stored.

set_params(fit_kwargs=None, compile_kwargs=None, custom_objects=None)

Set the parameters of the deep learning wrappers :param fit_kwargs: keyword arguments for the fit method. :param compile_kwargs: keyword arguments for the compile methods. :param custom_objects: This dict contains all custom objects needed by the keras model. Note,

users that uses such customs objects (e.g. Custom Loss) need to specify this to enable the loading of the stored Keras model.
transform(**kwargs) → xarray.core.dataarray.DataArray

Calls predict of the underlying keras Model. :param x: The dataset for which a prediction should be performed :return: The prediction. Each output of the keras model is a separate data variable in the returned xarray.

pywatts.modules.wrappers.pytorch_wrapper module

class pywatts.modules.wrappers.pytorch_wrapper.PyTorchWrapper(model: torch.nn.modules.module.Module, optimizer, loss_fn, name: str = 'PyTorchWrapper', fit_kwargs=None)

Bases: pywatts.modules.wrappers.dl_wrapper.DlWrapper

Wrapper for PyTorich Models.

Parameters:
  • model (torch.nn.Module,) – The pytorch model
  • name (str) – The name of the wrappers
  • fit_kwargs (Optional[Dict]) – Key word arguments used for fitting the model.
  • loss_fn (Callable) – The loss function of the model.
  • optimizer (Object) – The optimizer of the model.
fit(**kwargs)

Calls the compile and the fit method of the wrapped pytorch module.

get_params() → Dict[str, object]

Returns the parameters of deep learning frameworks. :return: A dict containing the fit keyword arguments and the compile keyword arguments

classmethod load(load_information) → pywatts.modules.wrappers.pytorch_wrapper.PyTorchWrapper

Method for restoring a pytorch Wrapper.

Parameters:load_information (Dict) – Dict which contains the information for restoring the wrappers
Returns:The restored wrappers
Return type:PyTorchWrapper
save(fm: pywatts.core.filemanager.FileManager)

Saves the pytorch wrappers and the containing model :param fm: Filemanager for getting the path :type fm: FileManager :return: Dictionary with additional information :rtype: Dict

set_params(fit_kwargs=None, loss_fn=None, optimizer=None)

Set the parameters of the deep learning wrappers :param fit_kwargs: keyword arguments for the fit method. :param compile_kwargs: keyword arguments for the compile methods.

transform(**kwargs) → xarray.core.dataarray.DataArray

Calls predict of the underlying PyTorch model.

Parameters:x – The dataset for which a prediction should be performed
Returns:The prediction. Each output of the PyTorch model is a separate data variable in the returned xarray.
class pywatts.modules.wrappers.pytorch_wrapper.TimeSeriesDataset(x, y)

Bases: torch.utils.data.dataset.Dataset

pywatts.modules.wrappers.sklearn_wrapper module

class pywatts.modules.wrappers.sklearn_wrapper.SKLearnWrapper(module: sklearn.base.BaseEstimator, name: str = None)

Bases: pywatts.modules.wrappers.base_wrapper.BaseWrapper

A wrappers class for sklearn modules. Should only used internal by the pipeline itself :param module: The sklearn module to wrap :type module: sklearn.base.BaseEstimator :param name: The name of the module :type name: str

fit(**kwargs)

Fit the sklearn module :param x: input data :param y: target data

get_params()

Return the parameter of the slkearn module :return:

inverse_transform(**kwargs) → xarray.core.dataarray.DataArray

Performs the inverse transform of a dataset with the wrapped sklearn module :param x: the input dataset :return: the transformed output

classmethod load(load_information) → pywatts.modules.wrappers.sklearn_wrapper.SKLearnWrapper
Parameters:load_information (Dict) – Information for reloading the SklearnWrapper
Returns:The reloaded SklearnWrapper
Return type:SKLearnWrapper

Warning

This method use pickle for loading the module. Note that this is not safe. Consequently, load only modules you trust. For more details about pickling see https://docs.python.org/3/library/pickle.html

predict_proba(**kwargs) → xarray.core.dataarray.DataArray

Performs the probabilistic transform of a dataset with the wrapped sklearn module :param x: the input dataset :return: the transformed output

save(fm: pywatts.core.filemanager.FileManager)

Saves the modules and the state of the module and returns a dictionary containing the relevant information.

Parameters:fm (FileManager) – the filemanager which can be used by the module for saving information about the module.
Returns:A dictionary containing the information needed for restoring the module

:rtype:Dict

set_params(**kwargs)

Set the parameter of the internal sklearn module :param kwargs: The parameter of the internal sklearn module :return:

transform(**kwargs) → xarray.core.dataarray.DataArray

Transforms a dataset or predicts the result with the wrapped sklearn module :param x: the input dataset :return: the transformed output

pywatts.modules.wrappers.sm_time_series_model_wrapper module

class pywatts.modules.wrappers.sm_time_series_model_wrapper.SmTimeSeriesModelWrapper(module: Type[statsmodels.tsa.base.tsa_model.TimeSeriesModel], name: str = None, module_kwargs=None, fit_kwargs=None, predict_kwargs=None, use_exog=True)

Bases: pywatts.modules.wrappers.base_wrapper.BaseWrapper

Wrapper for statsmodels modules. When adding this module to the pipeline, all inputs that starts with target are handled as endogenous variables and all other as exogenous variables.

Parameters:
  • module – The statsmodels module to wrap. Not this module should not be initialised.
  • name (str) – The name of the module
  • module_kwargs (dict) – The module keyword arguments necessary for creating the statsmodel module
  • fit_kwargs (dict) – The optional fit keyword arguments for fitting the model
  • predict_kwargs (dict) – The optional predict keyword arguments for predicting with the model (except start and end)
fit(**kwargs)

Fits the statsmodels module

Parameters:kwargs (xr.DataArray) – A dict of input arrays
get_params() → Dict[str, object]

Returns the parameters of the statsmodels module.

Returns:A dict containing the module keyword arguments, the fit keyword arguments, the predict keyword

arguments and the fitted model parameters :rtype: Dict

classmethod load(load_information) → pywatts.modules.wrappers.sm_time_series_model_wrapper.SmTimeSeriesModelWrapper

Loads a statsmodels wrappers

Parameters:load_information (Dict) – Information for reloading the StatsmodelsWrapper
Returns:The reloaded StatsmodelsWrapper
Return type:SmTimeSeriesModelWrapper

Warning

This method use pickle for loading the module. Note that this is not safe. Consequently, load only modules you trust. For more details about pickling see https://docs.python.org/3/library/pickle.html

save(fm: pywatts.core.filemanager.FileManager)

Saves the statsmodels wrappers and the containing model

Parameters:fm (FileManager) – FileManager for getting the path
Returns:Dictionary with all information for restoting the module
Return type:Dict
set_params(module_kwargs=None, fit_kwargs=None, predict_kwargs=None, use_exog=None)

Set the parameters of the statsmodels wrappers

Parameters:
  • module_kwargs (Dict) – keyword arguments for the statsmodel module.
  • fit_kwargs (Dict) – keyword arguments for the fit method.
  • predict_kwargs (Dict) – keyword arguments for the predict method.
transform(**kwargs) → xarray.core.dataarray.DataArray

Predicts the result with the wrapped statsmodels module

Parameters:kwargs (xr.DataArray) – A dict of input arrays
Returns:the transformed dataarray
Return type:xr.DataArray

Module contents