Metric Object

A metric is a variable that can be tracked in association with a Run. In an AI Experiment, a metric or set of metrics represents the performance of an associated parameter configuration.

KeyTypeValue

name

string

A user-specified name for a given metric

objective

string

Optional. Specifies the direction of improvement for a given metric. For any metric the objective can be to either “maximize” or “minimize”. If unspecified the default will be “maximize”.

strategy

string

Optional. Specifies how SigOpt will use a given metric. Can be “optimize”, “store”, or “constraints”. If unspecified, the metric will be optimized. Stored metrics can be helpful for tracking and additional analysis. See Using Metric Strategy for more details. Constraint metrics can be used to specify additional guardrail metrics. See Metric Constraints for more details.

threshold

float

The minimum or maximum desired value for a given metric. To apply thresholds on optimized metrics, see the Metric Thresholds documentation for more details. To apply thresholds on non optimized metrics, see the Metric Constraints documentation for more details. This field must be defined when the strategy is set to constraint.

Last updated