​
Overview

The BaseGlobalMetric class provides an abstract base for implementing global metrics in prompt evaluation. Global metrics are used to compute an overall score based on the results of individual metric evaluations.

​
Methods

compute
method

Abstract method to compute the global metric.

Parameters:

  • results (List[MetricResult]): The results from BaseMetric evaluations.

Returns: GlobalMetricResult: The computed global metric value.

__call__
method

Unified method to compute the global metric, handling both sync and async implementations.

Parameters:

  • results (List[MetricResult]): The results from BaseMetric evaluations. Use results[i].metadata to access the local metric results.

Returns: GlobalMetricResult: The computed global metric value.

​
Usage

To use the BaseGlobalMetric class, create a subclass and implement the compute method:

from ape.common.global_metrics import BaseGlobalMetric

class MyGlobalMetric(BaseGlobalMetric):
    def compute(self, results: List[MetricResult]) -> GlobalMetricResult:
        # Implement your global metric computation here
        return GlobalMetricResult(score=0.5)

Was this page helpful?

CosineSimilarityMetricAverageGlobalMetric