Polyaxon v1.7: Improved Assets Tracking, New Callbacks, Better Documentation with Examples, CSV Download, Better UI for Polyaxon Manifests, Improved Legends, Hover, and Tooltips for Plots, & Compatibility with Kubernetes 1.19

Today, we are pleased to announce the v1.7 release of our MLOps platform, a stable version that brings several new features, enhancements, and fixes. This release does not introduce any breaking changes and is fully compatible with previous releases.

Improved assets Tracking

Tracking assets and lineage information is an important aspect of machine learning reproducibility. In previous versions, users were often confused about the behavior of assets and artifacts tracking.

Polyaxon features two interfaces for tracking artifacts:

  • Versioned assets tracking: useful when the user needs to save the same asset but several times in the same run.
  • Reference tracking: gives users more control over where the asset must be saved.

We improved the interface and now have a documentation guide that explains the behavior and some use cases.

New ML/DL Callbacks

Callbacks and loggers allow an automated process for tracking params, metrics, charts, and assets. In v1.7, Polyaxon provides, several new callbacks for major deep learning and machine learning libraries and frameworks:

As well as several visualization libraries.


Better Documentation with Examples

We improved several aspects about the documentations:

The examples repo was also updated with several corrections and new samples.

Download as CSV in CLI and UI

Polyaxon now allows to download CSV results both using the CLI and the UI:

  • CLI:
polyaxon ops ls -io --limit=1000 --to-csv

A more complicated example:

polyaxon ops ls -io -c finished_at,run_time,status,learning_rate,loss,accuracy -q "status:succeeded" -s loss --limit 2000 --offset=1000 --to-csv
  • UI:



Better UI for Polyaxon Manifests

If you have large Polyaxonfiles, with several inputs, outputs, initializers, … you can easily fold sections in the UI:


Improved Legends, Hover, and Tooltips for Plots

The charts have several new capabilities to improve hovering effect and legend in mosaic mode:


And tooltips in scatter plots show more context:


We also fixed some minor issues in the charts when the dark theme is enabled.

Compatibility with Kubernetes 1.19

The operator now supports Kubernetes 1.19, this change has no backward-incompatible impact on the current deployments. For clusters with Kubernetes versions <1.15, we recommend enabling the previous behavior:

  useCRDV1Beta1: true

Several CLI Improvements

Several users already creates an alias plx for Polyaxon CLI. The CLI now ships by default with both entry-points polyaxon and plx.

polyaxon --help


plx --help

The CLI has also some new improvements for parsing and validating string types for Polyaxonfiles and passed params.

And finally, We fixed the logs streaming duplication issue when operations are in the running phase.

Future work

There are a some PRs that did not make it to this release, that will be part of v1.8:

  • Graph view for DAG and Matrix operations, this will come as an additional view to the timeline that was released in v1.6.
  • Re-enabling table/csv widgets in the dashboards tab.
  • Several improvements to the charts and graphs widgets.
  • Faster comparison tables using virtualization.

Currently the Agent/Queue service provides features to:

  • route operations to clusters and namespaces
  • manage concurrency and parallelism
  • prioritize operations

We started prototyping a new feature in the Agent/Queue manager to throttle operations based on resources (CPU/GPU/Memory/…) that will work natively with distributed operations as well. This mechanism will be an additional process to help optimize your queues and maximize your cluster usage.

Learn More about Polyaxon

This blog post just goes over a couple of features that we shipped since our last product update, there are several other features and fixes that 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 layer on Kubernetes. We hope that these updates will improve your workflows and increase your productivity, and again, thank you for your continued feedback and support.

Subscribe to Polyaxon Blog

Get the latest posts delivered right to your inbox