All Products
Search
Document Center

Platform For AI:View experiment reports

Last Updated:Aug 30, 2024

This topic describes how to view experiment reports.

Prerequisites

  • An experiment is created and started, and experiment metrics are associated with the experiment. For more information, see Create an experiment.

  • A table is registered and metrics are created. For more information, see Manage metrics.

Go to the experiment report page

  1. Go to the ABTest page.

    1. Log on to the PAI console.

    2. In the left-side navigation pane, click Workspaces. On the Workspaces page, click the name of the workspace that you want to manage.

    3. In the left-side navigation pane, choose Model Deployment > PAI-ABTest.

  2. In the left-side navigation pane of the ABTest page, choose Experiment Management > Experiments.

  3. In the experiment list, find the experiment that you want to view and click View Report in the Actions column. The Experiment Reports tab on the details page of the experiment appears.

View experiment reports

View experiment reports in different dimensions

You can filter the content of the experiment reports by statistical period and time range.

  • By statistical period

    Statistical period

    Description

    Day: Average per Day

    The system extracts metric data by day based on the selected time range and calculates the average values. This allows you to view metric data of different experiment versions over the time period.image

    Day: Statistics per Day

    The system performs aggregate computing on metric data by day. You can view the daily change trends of metric data.image

    Hour: Average per Hour

    The system extracts metric data by hour based on the selected time range and calculates the average values.image

    Hour: Statistics per Hour

    The system performs aggregate computing on metric data by hour. You can view the hourly change trends of metric data.image

  • By time rangeimage

View experiment report details

An experiment report displays the core information of an experiment in a visualized chart, including the experiment date, experiment versions, version types (experiment version and control version), and related metrics. You can use the metric data of the control version as the basis to compare with the metric data of each experiment version and visualize the differences among all versions in a chart.

image