google / nitroml

NitroML is a modular, portable, and scalable model-quality benchmarking framework for Machine Learning and Automated Machine Learning (AutoML) pipelines.
Apache License 2.0
41 stars 6 forks source link

Store pipeline run specific properties in MLMD #31

Closed nikhil-dce closed 3 years ago

nikhil-dce commented 4 years ago

We need to store useful properties for each pipeline (such as started_at).

cweill commented 3 years ago

This is obsolete in the IR-based DagRunner world.