[frontend] No metrics shown in Run Output tab

Environment

• How did you deploy Kubeflow Pipelines (KFP)? I installed it with the full kubeflow deployment. Kubeflow 1.6, installed with juju
• KFP version: build version dev_local (this shows on bottom of KFP UI left sidenav)

Steps to reproduce

Running this minimal pipeline that just saves a metrics

from kfp.components import OutputPath, create_component_from_func
import kfp.dsl as dsl
import kfp


@dsl.pipeline(
    name='kubeflow-metric-demo')
def metric_pipeline():
    produce_metrics_op = create_component_from_func(
        produce_metrics,
        base_image='python:3.7',
        packages_to_install=[],
        output_component_file='component.yaml',
    )
    produce_metrics_op()


def produce_metrics(
    mlpipeline_metrics_path: OutputPath('Metrics')
):
    import json

    metrics = {
        'metrics': [{
            'name': 'accuracy',
            'numberValue':  0.9,
            'format': "RAW"
        }]
    }
    with open(mlpipeline_metrics_path, 'w') as f:
        json.dump(metrics, f)
    print("metrics path")
    print(mlpipeline_metrics_path)


if __name__ == "__main__":
    kfp.compiler.Compiler().compile(metric_pipeline, 'metric_pipeline.yaml')

Expected result

I expect the metrics to be shown in the “Run output” tab in the UI. However, nothing is there as you can see in the image.

However, the metrics did show up at the “output artifacts” section of the component

Materials and Reference

---

Impacted by this bug? Give it a 👍.

[Lejboelle]

This has been fixed in the recent patch release https://github.com/kubeflow/manifests/releases/tag/v1.6.1

[CheaterAdams]

Hi @Kokkini, does your comment regarding platform agnostic Kubeflow 1.6.1 imply the fix is provided in 1.6.1? I am still looking for a solution using KPF 1.8.16.