Create Run

Click “+ Create run” button then a window titled “Start a New Run” appears, containing several fields that need to be filled out:

Create Run

a. Run details, you will select the pipeline you intend to use in this section. You can either create a new pipeline or opt for an existing one within your Service Portal Cloudeka Deka MLOps by clicking the “Use this pipeline” button. After selecting the pipeline, choose its version and provide a name and description for the run you initiate. Additionally, you must specify the experiment with which the run will be associated. This information is essential for effectively identifying and organizing your runs.

b. Run Type, this field allows you to specify the type of run you want to execute. There are two main options.

1. One-off Run: This is a single, standalone execution of the pipeline. The run is initiated manually and does not repeat automatically.

2. Recurring Run: This option enables you to set up a recurring or scheduled execution of the pipeline. When selecting a Recurring Run, you need to configure the following.

   1. Trigger Type: You can choose between two trigger types:

      — Periodic: This will run the pipeline at a regular interval, such as every X minutes, hours, days, weeks, or months.

      — Cron: This allows you to define a custom cron schedule to trigger the pipeline runs.

   2. Maximum Concurrent Runs: This setting determines the maximum number of pipeline runs that can be executed concurrently.

   3. Start Date and End Date (Optional): You can optionally specify a start and end date for the recurring run. This allows you to schedule the pipeline to run only during a specific time period.

   4. Catch-up (Optional): Some recurring run configurations may include a "catch-up" feature. This ensures that if a scheduled run is missed (e.g., due to a system outage), the pipeline will automatically execute the missed runs when the system becomes available again.

c. Pipeline Root: Here, you will define the root directory for the pipeline. This is where the artifacts generated during the run will be stored. Specifying a clear and organized pipeline root helps in managing outputs and logs effectively.

d. Run Parameters: This section is where you can input any parameters that the pipeline requires for execution. These parameters can include hyperparameters for machine learning models, input data paths, or any other configuration settings that are necessary for the run.

By filling out these fields, you set the foundation for your experiment, ensuring that all necessary details are in place for a successful execution of your pipeline in the Portal Cloudeka Deka MLOps Service and you can press “Start” button.

Create New Runs

After successfully creating a “Runs”, you will be directed to the Recurring Run menu, which displays information related to the Recurring Run that has been executed if you selected “Recurring” as the Run Type. If you wish to create another Recurring Runs, you can refer to the documentation on the Recurring Runs page documentations.

Recurring Runs Menu

However, if you selected “One-off” as the Run Type, you can simply press the “Start” button without needing to fill out the Run Trigger section. You will then be taken to the details page of the experiment once it has been successfully created.

Create New Runs

After successfully completing the process, you will be redirected back to the Active submenu in the Runs menu.