[sdk] SlicedClassificationMetrics load_roc_readings error

[petethegreat]

Environment

• KFP version: N/A
• KFP SDK version: 2.3.0
• All dependencies version:

kfp                      2.3.0
kfp-pipeline-spec        0.2.2
kfp-server-api           2.0.2

Steps to reproduce

SlicedClassificationMetrics.load_roc_readings() gives an error. When called, it tries to invoke ClassificationMetrics.load_roc_readings() - this method does not exist, giving an error. The corresponding ClassificationMetrics method is log_roc_curve The order of the parameters (fpr, tpr, thresholds) are different - documentation for load_roc_readings suggests thresholds, tpr, fpr, but log_roc_curve wants fpr, tpr, thresholds

Also, as mentioned in #8117 - SlicedClassificationMetrics does not initialise _sliced_metrics itself, resulting in an error if this is not done manually.

Expected result

When calling SlicedClassificationMetrics.load_roc_readings, the ROC metrics should be loaded correctly without error

Materials and Reference

example to reproduce:

from kfp import dsl
from kfp import compiler

@dsl.component(base_image='python:3.11')
def slicedmetrics_component(
    evaluation_metrics: dsl.Output[dsl.SlicedClassificationMetrics]):

    evaluation_metrics._sliced_metrics = {}

    # dummy roc data
    fpr = [ 0.0, 0.0, 0.0, 1.0]
    tpr = [0.0, 0.5, 1.0, 1.0]
    thresholds = [float('inf'), 0.99, 0.8, 0.01]

    evaluation_metrics.load_roc_readings(
        'test_slice', [thresholds, tpr, fpr]
    )

@dsl.pipeline(
    name='slicedclassification_sample_pipeline')
def example_pipeline():
    metrics_op = slicedmetrics_component()

compiler.Compiler().compile(example_pipeline, package_path='slicedclassification_pipeline.json')

---

Impacted by this bug? Give it a 👍.

[zijianjoy]

/assign @connor-mccarthy

Possible guess: float('inf') blocks json serialization.

[petethegreat]

pr that addresses this is here: #10350

[github-actions[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[github-actions[bot]]

This issue has been automatically closed because it has not had recent activity. Please comment "/reopen" to reopen it.