Skip to content

0.23.0
Compare
Choose a tag to compare
@schustmi schustmi released this 02 Dec 13:45
· 1749 commits to main since this release
0.23.0
df35bf8
This commit was signed with the committer’s verified signature.
schustmi Michael Schuster
GPG key ID: C5884EDD066312AA
Learn about vigilant mode.

This release comes with a brand-new Neptune integration to track your ML experiments

as well as lots of performance improvements!

Neptune integration

The new Neptune integration
includes a Neptune experiment tracker component that allows you to track your machine learning experiments
using Neptune.

Performance Optimization

The 0.20.0 release introduced our new server but brought with it a few performance and scalability issues.
Since then we've made many improvements to it, and this release is the final and biggest boost in performance. We reduced the amount of server calls needed for almost all CLI commands and greatly improved the speed of the dashboard as well.

PyArrow dependency removal

We've removed PyArrow as a dependency of the zenml python package.
As a consequence of that, our NumPy and Pandas materializer no
longer read and write their artifacts using PyArrow but instead use
native formats instead. If you still want to use PyArrow to serialize
your NumPy arrays and Pandas dataframes, you'll need to install it manually
like this: pip install pyarrow

In future releases we'll get rid of other unnecessary dependencies to
further slim down the zenml package.

Breaking Changes

The following changes introduces with this release mey require some manual
intervention to update your current installations:

  • If your code calls some methods of our Client class, it might need to be
    updated to the new model classes introduced by the performance optimization changes
    explained above
  • The CLI command to remove an attribute from a stack component now takes no more dashes
    in front of the attribute names:
    zenml stack-component remove-attribute <COMPONENT_NAME> <ATTRIBUTE_NAME>
  • If you're using a custom stack component and have overridden the cleanup_step_run method,
    you'll need to update the method signature to include a step_failed parameter.

What's Changed

New Contributors

Full Changelog: 0.22.0...0.23.0

Contributors

jcarlosgarcia, htahir1, and 7 other contributors
Assets 2
Loading