NitroML is a modular, portable, and scalable model-quality benchmarking framework for Machine Learning and Automated Machine Learning (AutoML) pipelines.
Task should have problem_statement property which returns a ProblemStatement proto. Several TFX components consume this proto, and therefore we should use it.
A task is essentially a two-tuple of (dataset, problem_statement).
We only need to populate the minimal fields on the proto. A minimalist proto (in text format) would look like:
Task should have
problem_statement
property which returns aProblemStatement
proto. Several TFX components consume this proto, and therefore we should use it.A task is essentially a two-tuple of (dataset, problem_statement).
We only need to populate the minimal fields on the proto. A minimalist proto (in text format) would look like: