Implementation:Kubeflow Pipelines Regression Metrics Evaluation

Sources: Kubeflow Pipelines

Domains: Machine_Learning, Evaluation

Last Updated: 2026-02-13

Overview

Wrapper Doc for the reusable KFP component that calculates regression metrics from CSV predictions and ground truth.

Description

calculate_regression_metrics_from_csv_op is loaded from a remote YAML spec. It compares true values against predicted values and outputs multiple regression metrics including mean_squared_error.

Code Reference

Source: samples/core/train_until_good/train_until_good.py (L28 loading, L52-55 invocation)

Import: from kfp import components

calculate_regression_metrics_from_csv_op = components.load_component_from_url(
    'https://raw.githubusercontent.com/kubeflow/pipelines/.../components/ml_metrics/Calculate_regression_metrics/from_CSV/component.yaml'
)

metrics_task = calculate_regression_metrics_from_csv_op(
    true_values=true_values,
    predicted_values=predictions,
)

# Access outputs
mse = metrics_task.outputs['mean_squared_error']

I/O Contract

Inputs

Name	Type	Required	Description
true_values	CSV	Yes	Ground truth values
predicted_values	CSV	Yes	Model predictions

Outputs

Name	Type	Description
mean_squared_error	float	Mean squared error between predictions and ground truth
mean_absolute_error	float	Mean absolute error between predictions and ground truth
root_mean_squared_error	float	Root mean squared error between predictions and ground truth

Related Pages

• Principle:Kubeflow_Pipelines_Model_Evaluation_Metrics