Right now, all the Scalers and Feature Selectors and other things are defined as instances of their respective SKLearn classes, but also as an instance of a Base class that is and instance of Step.
This is because a Step object must implement some common functions and extract the parameters passed to the init function for later saving, thus they all get redefined and wrapped by VDB to make them into Steps
It would be nice if there was a way that, if the code detected something passed to a Pipeline/Workflow that wasn't an instance of Step, but was a BaseEstimator (from SKlearn) it would auto convert it into an instance of the Step class too. This might be impossible, but meta classes come to mind for solving it.
Right now, all the Scalers and Feature Selectors and other things are defined as instances of their respective SKLearn classes, but also as an instance of a Base class that is and instance of Step.
This is because a Step object must implement some common functions and extract the parameters passed to the init function for later saving, thus they all get redefined and wrapped by VDB to make them into Steps
It would be nice if there was a way that, if the code detected something passed to a Pipeline/Workflow that wasn't an instance of Step, but was a BaseEstimator (from SKlearn) it would auto convert it into an instance of the Step class too. This might be impossible, but meta classes come to mind for solving it.
Any ideas?