Customizable Tuners

Polyaxon has a built-in serverless optimization engine; the way Polyaxon schedules a hyperparameter tuning pipeline is by creating suggestions using a tuner. The tuner is the component responsible to query previous observations and suggesting new configurations to explore.

In this release, we are making all built-in tuners customizable, we just moved all tuners to the public component hub. A user can just provide a custom component to change the default tuning behavior.

For instance to override the default Hyperband tuner you just need to provide your own component instead of the default implementation.

Finally, the scheduler is now using the joins abstraction to query the historical observations, not only this will automatically create upstream/downstream edges between the tuner and the observations, but also it will help with debugging the behavior of each iteration.

Customizable Notifiers

Similar to the tuners, Polyaxon now provides fully customizable notifiers and they are all hosted on the public component hub.

Since v1.5, Polyaxon moved the default notifications logic to hooks. Hooks are an automation feature and a more customizable interface that execute post-done operations based on a component reference, for example, a user can both send a slack notification and start a tensorboard after their experiment is successful:

  - trigger: succeeded
    hubRef: tensorboard
    disableDefaults: true
      uuid: { value: '{{ globals.uuid }}' }
  - trigger: succeeded
    hubRef: slack
    connection: slack-notification

The current notifiers’ implementation was also improved and includes several more information, like the inputs and outputs:


Users can easily clone the default implementation and add custom information or charts to send to their favorite channel.

Learn More about Polyaxon

This blog post just goes over a couple of features that we shipped since our last product update, several other features and fixes are worth checking. To learn more about all the features, fixes, and enhancements, please visit the release notes.

Polyaxon continues to grow quickly and keeps improving and providing the simplest machine learning abstraction. We hope that these updates will improve your workflows and increase your productivity, and again, thank you for your continued feedback and support.