ML Jobs#

Introduction#

Machine Learning Jobs is used to manage the machine learning jobs associated with all Machine Learning Workflows in a project.

Usage#

In the Project Settings tab select ML Jobs. This will list all the Machine Learning jobs for all the Machine Learning Workflows in a project as shown below. A new job can be created by clicking on the plus button on the top right corner of the screen.

image1

The list view provides a summarised at-a-glance view for each job with relevant details like the current status of the job, total number of runs, and next run time.

Tip: More details for a particular job can be accessed by clicking on the row for that particular job in the list view. It will open a more detailed view for each job as shown below.

image2

The most notable piece of information in the detailed view is the Last result section, which gives you a summarised view of the training accuracy of the model as in how many labels are correctly/wrongly classified by the model for each class.

Note: Since it is a view on the training accuracy, this section is only populated for jobs of type training and NOT for jobs of type inference.

By clicking Show Logs, you can also access the run time logs of that particular job, which is often useful in debugging a failed/erred job or to understand (and improve) the quality for a particular model during the experimentation phase by being able to inspect what the model is doing.

Killing and Deleting Jobs#

In addition to being able to inspect all kinds of metadata around a particular job, the plugin also provides two useful actions for each job:

  • Kill Job: Kills a running Machine Learning job. This is often useful if a particular job is taking too much of your CPU resources and affecting the performance of your Squirro server.

  • Delete Job: Delete a Machine Learning job