Skip to content

Releases: opendatahub-io/odh-dashboard

v2.21.1-incubation

08 Mar 15:13
Compare
Choose a tag to compare

Details

Moving forward we are reducing our overhead and doing away with the incubation concept. As we move forward we will just be "early releasing" our code into ODH based on what we have actively being developed in our main branch. This has some added benefits of making one stream of code and fixes so there is less "how do we get this for ODH vs RHOAI". This also has the added benefit of working to get all our early access work on RHOAI features to ODH as we develop them

We are at the same time looking to do away with our feature branches -- we will clean these up in the near future, everything should be locked behind feature flags (which will also be getting a facelift soon)

If you have any questions about this process, feel free to log a question ticket and we'll get back to you asap

It should be noted, there is a lossy effect with this release.

  • We dropped Service Mesh support & the feature flag
  • We may have lost some small feature that we were unaware of
  • This is a reset on main with community features brought over
  • If you have any knowledge of a loss of a feature and want it back, please let us know with a question ticket

What's Changed

New Contributors

Full Changelog: v2.21.0...v2.21.1-incubation

v2.21.0

04 Mar 13:31
bab4396
Compare
Choose a tag to compare

Noteable Changes

  • A workaround for the lack of a default storage class -- using the spec.notebookController.storageClassName field in the OdhDashbaordConfig will allow you to set one for the DS Projects (only if there is no default storage class -- does nothing for DS Projects if you do have one)
  • Initial Distributed Workloads work (disabled by default) - more will be said the further along we get
  • Enable GRPC support for model serving routes
  • Misc bug fixes & improvements

What's Changed

New Contributors

Full Changelog: v2.20.0...v2.21.0

v2.20.0

09 Feb 21:45
ea16972
Compare
Choose a tag to compare

Notable Changes

  • Pipeline features
    • Pipeline Logs
      • The ability to view the logs behind each step of a task in the details page of Pipeline Runs
    • Pipeline Versions
      • The support for Pipeline Versions in the Pipelines page (this replaced the pre-existing last 5 runs)
      • You should be able to filter on runs associated with each version
    • Pipeline resources are now entirely server-side manageed
      • This has changed the filtering and sorting mechanisms of the various tables
      • You can now paginate through all the various resources to see all your data (largely this will improve viewing runs)
    • We are working on release v2 of Pipelines, and thus there is now a deprecation notice on v1
      • After you update to the v2 release (aimed at next release) your Pipeline UI will no longer be usable inside the Dashboard due to incompatibility in the API versions
      • API is still directly available (until you delete your DSPA)
      • Elyra notebook usage is still available (until you delete your DSPA)
  • Other misc fixes & improvements

What's Changed

Read more

v2.19.1-incubation

01 Feb 18:32
Compare
Choose a tag to compare

Features Incubating

These are based on the latest release to main (Release Notes for v2.19.0) + some bug fixes fixes

  • f/storage-class: A workaround for not having a default Storage Class set on your cluster
  • f/ossm: Initial implementation of Service Mesh & Istio Proxy
  • f/pipelines-enhancements: Pipelines Logs, Pipelines Versions & other improvements for DS Pipelines (eg. server side paging requests)

What's Incubation?

Read more about incubation in our docs.

If you have any questions regarding incubation, please log a question-based issue. We look forward to improving the flow and adding more clarity.

v2.19.0

19 Jan 18:23
f2de311
Compare
Choose a tag to compare

Notable Changes

  • Accelerator Admin page support
  • Model Severing metrics
    • TrustyAI Bias detection
    • Model & Server Performance metrics
    • User Workload Monitoring (UWM) support
  • Upgrade to our Patternfly component library to v5; slight changes to visuals may be seen
  • Misc bug fixes

What's Changed

Read more

v2.18.1-incubation

07 Dec 14:11
Compare
Choose a tag to compare

Features Incubating

These are based on the latest release to main (Release Notes for v2.18.0) + some bug fixes fixes

  • Reworks the incubation branch to include our component library upversion, Patternfly (v5)
  • f/storage-class: A workaround for not having a default Storage Class set on your cluster
  • f/ossm: Initial implementation of Service Mesh & Istio Proxy
  • f/mserving-metrics: Additional features for Model Serving - TrustyAI Bias, Performance Metrics, UWM support
  • f/accelerator-admin-support: Admin flows for adding Accelerator Profiles
  • f/pipelines-enhancements: Logs & other improvements for DS Pipelines

What's Incubation?

Read more about incubation in our docs.

If you have any questions regarding incubation, please log a question-based issue. We look forward to improving the flow and adding more clarity.

v2.18.0

17 Nov 19:28
e8b1c06
Compare
Choose a tag to compare

Notable Changes

  • OpenShift Console & DS Projects Improvements
    • Projects created in OpenShift Console can now be read in by the Dashboard
    • There is a dropdown on the DS Projects list view that allows you to switch between DS only Projects and all projects your user has RBAC access to
    • There is a limiting nature to some of the projects shown here, but they should only impact the cluster-admin user (you will not be able to see the openshift- and kube- prefixed projects that you can in OpenShift Console, for instance)
    • We remember this setting for convenience
    • Data Science flows should work fine in these projects
  • KServe feature in the Operator's DataScienceCluster (DSC) is now available in the Dashboard
    • Single model (KServe) & multi-model (Model Mesh) are now supported in the Dashboard
    • New installs will favour single model Model Serving
    • Existing installs should continue running on the multi-model Model Serving
  • "Area" Checks now auto hide Dashboard flows
    • If you disable a resource in the DSC (managementState: Removed), it should hide the corresponding section in the UI
    • Feature flags exist for the same set of DSC resources (in addition to several flows specific to the UI -- like Projects View)
      • This is intended to give extra flexibility in hiding UI flows but keeping on-cluster resources for GitOps or other API driven flows
    • Disabling either DSC managementState status or the feature flag will hide that area in the Dashboard
      • If the area is reliant on another area that is disabled, the flows for that area should hide as well
    • For more details on how the relationship is constructed, see our Areas map
  • General bug improvements

What's Changed

New Contributors

Full Changelog: v2.17.0...v2.18.0

v2.17.0-incubation

27 Oct 14:37
Compare
Choose a tag to compare

Features Incubating

These are based on the latest release to main (Release Notes)

  • Model Serving Metrics - #1128
  • Model Bias/Explainability - #1127
  • Pipeline Experiments (Modals / Table / Run creation flows) - #1115
  • Pipeline Enhancements - #1651
    • Logs
    • Better param handling
    • Duplicating a run copies last run input parameters
  • Default-less Storage Class Clusters for DS Projects - #1919

    Edge case support for no storage class default on your cluster
    Make use of this by using the spec.notebookController.storageClassName field

What's Incubation?

Read more about incubation in our docs.

If you have any questions regarding incubation, please log a question-based issue. We look forward to improving the flow and adding more clarity.

v2.17.0

27 Oct 14:33
265b0b2
Compare
Choose a tag to compare

Notable Changes

  • Improved our Admin Custom Notebook Image flow
    • Should be easier to find your resource now in OpenShift Console -- a popover is next to the image name with a copy feature
    • There is better detection on create
    • Improved the software and package flows when adding metadata to your images
  • Bug fixes

What's Changed

Full Changelog: v2.16.0...v2.17.0

v2.16.0

06 Oct 14:08
b9f4a84
Compare
Choose a tag to compare

Notable Changes

  • AcceleratorProfiles (and Habana support) are now available
    • Accelerators are the group term for GPUs & AI specific chips that don’t include the Graphics (“G” in GPU) aspect
    • An AcceleratorProfile is a way to describe the backing resource
    • AcceleratorProfiles have “identifier” to speak nvidia.com/gpu or any brand that a Node would understand and be able to schedule
      • Each profile has the ability to come with a description & and display name to curate it for your users
      • Tolerations can not be specified more openly for existing Nvidia GPU and any future accelerators you want to support
    • Since this is the grouping term for our existing support of Nvidia GPUs, a couple things are updated in this area
      • On upgrade, if you have GPUs installed, we should detect that and create you a migrated-gpu AcceleratorProfile
      • If you have any problems with this, an AcceleratorProfile can be created in the application namespace (where the Dashboard is deployed)
    • We welcome all feedback, please log an issue if you see gaps in our flow for your use-cases
  • Bug fixes

What's Changed

New Contributors

Full Changelog: v2.15.0...v2.16.0