The world's cleanest AutoML library ✨ - Do hyperparameter tuning with the right pipeline abstractions to write clean deep learning production pipelines. Let your pipeline steps have hyperparameter spaces. Design steps in your pipeline like components. Compatible with Scikit-Learn, TensorFlow, and most other libraries, frameworks and MLOps environments.
Sometimes, I need to fetch all steps within a pipeline which are instances of a given class (e.g. some featurizer class which I use in multiple place). Usually, this happen when I want compute something based on the settings of all instances of this class in the pipeline (e.g. each of the Featurizer have a "dimension" attribute).
So far, I manage this by creating a function with a unique name, then use the method apply(method_name) on the root of the pipeline to retrieve the information/the steps I want. This workaround seems problematic to me for many reason. Mainly:
it restrict the space of method name we can use when writing new steps (i.e. if you need that method name to be unique, then you potentially need to be aware of ALL method name of ALL steps used within your pipeline, which isn't realistic if your pipeline is massive and you have code it all by yourself)
Frivolous use of apply creates RecursiveDict which we don't really need to instantiate.
To fix this, we should add a method, like get_step_by_name, which returns all steps which are instance of a given class.
This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs in the next 180 days. Thank you for your contributions.
Sometimes, I need to fetch all steps within a pipeline which are instances of a given class (e.g. some featurizer class which I use in multiple place). Usually, this happen when I want compute something based on the settings of all instances of this class in the pipeline (e.g. each of the Featurizer have a "dimension" attribute).
So far, I manage this by creating a function with a unique name, then use the method apply(method_name) on the root of the pipeline to retrieve the information/the steps I want. This workaround seems problematic to me for many reason. Mainly:
To fix this, we should add a method, like get_step_by_name, which returns all steps which are instance of a given class.