Skip to content

Release 0.8.20

Need Help?

The SuperCowPowers team is happy to give any assistance needed when setting up AWS and Workbench. So please contact us at workbench@supercowpowers.com or on chat us up on Discord

The Workbench framework continues to flex to support different real world use cases when operating a set of production machine learning pipelines.

Note: These release notes cover the changes from 0.8.11 to 0.8.20

General

The cloud_watch AWS log aggregator, is now officially awesome. It provides a fairly sophisticated way of both doing broad scanning and deep dives on individual streams. Please see our Cloud Watch documentation.

The View classes have finished their refactoring. The 'read' class View() can be constructed either directly or with the ds/fs.view("display") methods. See Views for more details. There also a set of classes for constructing views, please see View Overview

API Changes

  • Table Name attribute

    The table_name attribute/property has been replaced with just table

    ds.table_name -> ds.table
fs.table_name -> fs.table
view.table_name -> view.table

  • Endpoint Confusion Matrix

    The endpoint class had a method called confusion_matrix() this has been renamed to the more descriptive generate_confusion_matrix(). Note: The model method, of the same name, has NOT changed.

    end.confusion_matrix() -> end.generate_confusion_matrix()
model.confusion_matrix() == no change

Bug Fixes

Fixed: There was a corner case where if you had the following sequence:

  • New FeatureSet without a training view
  • Created a traiing view (like with set_training_holdouts()
  • Immediately query/pull from the FS training view

The corner case was a race-condition where the FeatureSet would not 'know' that a training view was already there and would create a default training view.

Improvements

The log messages that you receive on a plugin validation failure should now be more distinguishable and more informative. They will look like this and in some cases even tell you the line to look at.

ERROR Plugin 'MyPlugin' failed validation:
ERROR    File: ../workbench_plugins/components/my_plugin.py
ERROR    Class: MyPlugin
ERROR    Details: my_plugin.py (line 35): Incorrect return type for update_properties (expected list, got Figure)

Internal API Changes

In theory these API should not affect end user of the Workbench API but are documented here for completeness.

  • Artifact Name Validation

The internal method used by Artifact subclasses has changed names from ensure_valid_name to is_name_valid, we've also introduced an optional argument to turn on/off lowercase enforcement, this will be used later when we support uppercase for models, endpoints, and graphs.

Questions?

The SuperCowPowers team is happy to answer any questions you may have about AWS and Workbench. Please contact us at workbench@supercowpowers.com or on chat us up on Discord