Model Interface

BaseModel(model=None, change=False, model_path='', backend='Func')

Bases: ABC

Initialize Base Model. Arguments: model: trained ML Model, either the model or the direct function call for returning the probability distribution. change bool: True if dimension change is necessary. model_path str: path to trained model. backend str: ML framework. For frameworks other than TensorFlow (TF), Sklearn (SK) or PyTorch (PYT), provide 'Func'.

load_model(path) abstractmethod

Loads the model provided at the given path. Arguments: path str: Path to the trained model- Returns: loaded model.

predict(item) abstractmethod

Unified prediction function. Arguments: item np.array: item to be classified Returns: an array of output scores for a classifier, and a singleton array of predicted value for a regressor.

TensorFlowModel(model, change=False)

Bases: BaseModel

Wrapper for Tensorflow Models that unifiy the prediction function for a classifier. Arguments: model : Trained TF Model. change bool: if swapping of dimension is necessary = True

predict(item)

Unified prediction function. Arguments: item np.array: item to be classified. Returns: an array of output scores for a classifier.

Module containing an interface to trained PyTorch model.

PyTorchModel(model, change=False)

Bases: BaseModel

Wrapper for PyTorch Models that unifiy the prediction function for a classifier. Arguments: model : Trained PYT Model. change bool: if swapping of dimension is necessary = True

predict(item)

Unified prediction function. Arguments: item np.array: item to be classified. Returns: an array of output scores for a classifier.

SklearnModel(model, change=False)

Bases: BaseModel

Wrapper for Sklearn Models that unifiy the prediction function for a classifier. Arguments: model : Trained Sklearn Model. change bool: if swapping of dimension is necessary = True

predict(item)

Unified prediction function. Arguments: item np.array: item to be classified. Returns: an array of output scores for a classifier.