experiment

experiment

Reproducible experiment runs, metrics, datasets, models, and artifacts.

Classes

Name Description
LocalTracker Create runs in QRT’s local DuckDB and artifact store.
MLflowTracker Create runs in a configured MLflow tracking server or file store.
Run A context-managed experiment run with backend-neutral logging.

LocalTracker

experiment.LocalTracker(path='.qrt/experiments')

Create runs in QRT’s local DuckDB and artifact store.

MLflowTracker

experiment.MLflowTracker(experiment, tracking_uri=None, registry_uri=None)

Create runs in a configured MLflow tracking server or file store.

Run

experiment.Run(backend, *, name, tags)

A context-managed experiment run with backend-neutral logging.

Methods

Name Description
log_record Log a backend-neutral structured record as metadata or an artifact.
log_record
experiment.Run.log_record(kind, name, record)

Log a backend-neutral structured record as metadata or an artifact.

Functions

Name Description
active_run Return the run active in the current context, if any.
default_tracker Return the lazily constructed process-local tracker.
run Create a local run unless an explicit tracker is supplied.

active_run

experiment.active_run()

Return the run active in the current context, if any.

default_tracker

experiment.default_tracker()

Return the lazily constructed process-local tracker.

run

experiment.run(name, *, tracker=None, tags=None)

Create a local run unless an explicit tracker is supplied.

Back to top