Good to have a more robust way of tracking ML experiments, runs and model's we're using across variety of projects. We can use MLFlow to do so. MLFlow is created by databricks, can be managed by databricks, but can also be used without a databricks instance. Ex we can have our artifact store be a GCP bucket in our GCP project and additionally have a postgres database set up for tracking experiment, run, metadata.
Good to have a more robust way of tracking ML experiments, runs and model's we're using across variety of projects. We can use MLFlow to do so. MLFlow is created by databricks, can be managed by databricks, but can also be used without a databricks instance. Ex we can have our artifact store be a GCP bucket in our GCP project and additionally have a postgres database set up for tracking experiment, run, metadata.