MLBazaar / MLBlocks

A library for composing end-to-end tunable machine learning pipelines.
https://mlbazaar.github.io/MLBlocks
MIT License
114 stars 35 forks source link

Expose required init_params #119

Open csala opened 4 years ago

csala commented 4 years ago

Some primitives require the user to provide some init_params, and at the moment the list of these required init_params is only available in the primitive specification.

This can make using a pipeline complicated, as the only way to figure out what arguments needed to create an instance is to read the pipeline JSON, see which primitives the pipeline is using, and then load each all the primitives to see what the required init_params are for each one of them.

This issue is to make an MLPipeline classmethod which allows exploring the required init_params of a pipeline before actually loading it.