Release Notes

• 1: 2023.3.0 Product Release Notes (Beta Release)
• 2: 2023.2.1 Product Release Notes
• 3: 2023.2 Product Release Notes
• 4: 2023.1 Product Release Notes
• 5: 2022.4 (Dec 2022) Product Release Notes
• 6: September 2022 Product Release Notes
• 7: August 2022 Product Release Notes
• 8: July 2022 Product Release Notes
• 9: June 2022 Product Release Notes
• 10: May 2022 Product Release Notes
• 11: April 2022 Product Release Notes
• 12: March 2022 Product Release Notes
• 13: February 2022 Product Release Notes

1 - 2023.3.0 Product Release Notes (Beta Release)

2023.3.0 Release Overview

We are pleased to announce the following product improvements in our 2023.3.0 release:

• Edge Deployment: Wallaroo pipelines and pipeline deployment configurations can be published to Open Container (OCI) Registry services set in the Wallaroo configuration as the Edge Registry service. This allows organizations to retrieve published Wallaroo pipelines and deploy them to Edge devices through Docker, Docker Compose, Helm through Kubernetes, or other OCI compliant systems. See the reference documentation for more details:

  • Edge Deployment Registry Guide.
  • Wallaroo MLOps API Essentials Guide: Pipeline Edge Publishing Management.
  • Wallaroo SDK Essentials Guide: Pipeline Edge Publication.
  • Wallaroo Edge Deployment Tutorials

• Ampere Arm Architecture Support: Wallaroo pipelines can be deployed to x86 or ARM supported architectures, such as the as Ampere® Altra® Arm-based processor included with the following Azure virtual machines:

  • Dpsv5 and Dpdsv5-series
  • Epsv5 and Epdsv5-series

  This allows organizations to leverage the ARM architecture for energy and cost savings with similar. Using the Ampere AIO Accelerator enables even more efficiency and performance.

  See the following reference documentation for more details.

  • Create ARM Nodepools for Kubernetes Clusters.
  • Wallaroo MLOps API Essentials Guide: Model Upload and Registrations.
  • Wallaroo SDK Essentials Guide: Model Uploads and Registrations.
  • Wallaroo Features Tutorials: Wallaroo Pipeline Architecture.

2 - 2023.2.1 Product Release Notes

2023.2.1 Release Overview

We are pleased to announce the following product improvements in our 2023.2.1 release:

• Model Upload Expansion: With this release, Wallaroo expands the number of supported machine learning frameworks. ML Models that meet the supported model frameworks can be uploaded and deployed into Wallaroo workspaces and pipelines from an array of different model types and flavors.
  • Supported frameworks include the following. See the reference documentation for full framework and version requirements.
    • ONNX
    • Python Shape Steps
    • Tensorflow
    • Keras
    • Arbitrary Python
    • Hugging Face
    • PyTorch
    • Sci-kit Learn aka SKLearn
    • XGBoost
    • Containerized MLFlow
  • References:
    • Wallaroo SDK Essentials Guide: Model Uploads and Registrations: Guides on uploading ML models to a Wallaroo instance using the Wallaroo SDK.
    • Wallaroo MLOps API Essentials Guide: Model Upload and Registrations: Guides on uploading ML models to a Wallaroo instance using the Wallaroo MLOps API.
    • Wallaroo ML Models Upload and Registrations Guides: Tutorials on uploading, registering, and deploying ML Models of various frameworks into a Wallaroo instance.
• GPU Pipeline Deployment Support: Pipeline configuration includes allocating GPUs to pipeline deployment configurations both native and containerized runtimes.
  • References:
    • Create GPU Nodepools for Kubernetes Clusters
    • Wallaroo MLOps API Essentials Guide: Pipeline Management
    • Wallaroo SDK Essentials Guide: Pipeline Deployment Configuration
    • GPU Pipeline Deployment Tutorial
• MLFlow Registry Integration: Model Registry services connection details are made available to workspace users through the Wallaroo registry object. This provides methods of listing available models and their versions in a model registry, then uploading supported model artifacts into Wallaroo for pipeline deployments.
  • References:
    • Wallaroo SDK Essentials Guide: Model Uploads and Registrations: Model Registry Services
    • Wallaroo MLOps API Essentials Guide: Model Registry
    • Model Registry Service with Wallaroo SDK Demonstration
• Pipeline Log Reliability Enhancements: Wallaroo Pipeline Logs received the following updates.
  • Python shape logging: Python scripts uploaded as ML Models that are constructed to accept pandas DataFrame or Apache Arrow inputs and outputs will have the same inference result and logging structures as other ML Models deployed in Wallaroo. This places the inputs and outputs under the in and out metadata structures for consistent data input and retrieval.
  • Pipeline log reliability: Pipeline logs have a set storage space. To prevent log storage issues, inference result data may be restricted depending on the size of the input and output data. Inference results will always include all input and output details, while pipeline logs may drop extraneous data for storage purposes. The new update provides methods for developers to detect that information may be dropped from log storage and plan accordingly.
  • References:
    • Wallaroo SDK Essentials Guide: Pipeline Log Management
    • Wallaroo MLOps API Essentials Guide: Pipeline Log Management
• Arbitrary Python Model Support: Wallaroo ML Model support includes Arbitrary Python, providing organizations the ability to Python scripts with artifacts and libraries as ML models for deployment as Wallaroo pipeline steps. This provides a powerful level of flexibility in deploying models in a Wallaroo environment with additional support contained around them. Note that ML models included with an Arbitrary Python model must comply with Wallaroo model framework and versions requirements.
  • References:
    • Wallaroo SDK Essentials Guide: Model Uploads and Registrations: Arbitrary Python
    • Wallaroo MLOps API Essentials Guide: Model Upload and Registrations
    • Wallaroo SDK Upload Arbitrary Python Tutorials

3 - 2023.2 Product Release Notes

2023.2 Release Overview

We are pleased to announce the following product improvements in our 2023.2 release:

• ML Workload Orchestration: Wallaroo platform users (data scientists or ML Engineers) have the ability to deploy, automate and scale recurring production ML workloads that can ingest data from predefined data sources to run inferences in Wallaroo, chain pipelines, and send inference results to predefined destinations to analyze model insights and assess business outcomes. The following resources are available:.
  • ML Workload Orchestration Configuration Guide to enable orchestrations in a Wallaroo 2023.2 instance.
  • Wallaroo SDK Essentials Guide: ML Workload Orchestration for instructions on orchestration requirements, how to upload them and schedule them for use.
  • ML Workload Orchestration Tutorials: An expanding list of tutorials showing how to automate processes and integrations with systems such as Google Big Query, etc.
• Default Apache Arrow support: Apache arrow is now enabled by default for inference requests and results. Inference requests are submitted as pandas DataFrames, Apache Arrow tables, or custom JSON. Inference requests are sent via the Wallaroo SDK or through the Wallaroo API.
• Wallaroo Data Connections: MLOps Engineers can define data connection configurations, then relate them to Wallaroo workspaces for others users to access. This allows organizations to define connections to data stores, then make them available to other users to automate data retrieval and storage needs during ML operations. See the following guides for details:
  • Wallaroo SDK Essentials Guide: Data Connections Management
  • Wallaroo MLOps API Essentials Guide: Connections Management
• Drift Detection Assays on multiple inputs and outputs: Data scientists in Wallaroo can run drift detection assays to monitor specific model inputs or outputs, with their specific fields and their indexes for ML pipelines deployed in Wallaroo. For more details, see the following:
  • Wallaroo Assays Management
  • Model Insights Tutorials
  • Wallaroo SDK Essentials Guide: Assays Management
  • Wallaroo MLOps API Essentials Guide: Assays Management

4 - 2023.1 Product Release Notes

2023.1 Release Overview

As we continue to iterate on our core capabilities, we are pleased to announce the following product improvements in our 2023.1:

• Arrow and DataFrame Inference Input Support: Inferences can be run with inputs either from the Wallaroo proprietary JSON format, Pandas DataFrame, or Apache Arrow inputs, which return the same outputs in kind. This is an optional update; please contact your Wallaroo support representative on how to update your Wallaroo instance to enable Arrow support.
• Single Node Linux support for Wallaroo Enterprise Installations: Wallaroo Enterprise users can use our Single Node Linux installation guide to host a Wallaroo instance. Our updated guide provides sample templates for major cloud providers to help organizations get Wallaroo running even faster. This is an optional method of installing Wallaroo enterprise.
• Azure Databricks support: Organizations can leverage their Azure Databricks workspaces and deploy their models to a Wallaroo instance, including using Microsoft Azure as an identity provider for authentication.
• Amazon Sagemaker Support: AWS Sagemaker users can use the Wallaroo SDK to connect with their Wallaroo instance to deploy models, perform inferences, and other vital tasks. Support for Single Sign On using Amazon AWS allows for seamless authentication across AWS services.

5 - 2022.4 (Dec 2022) Product Release Notes

2022.4 Release Overview

With the 2022.4 release, Wallaroo shifts into a quarterly release schedule.

As we continue to iterate on our core capabilities, we are pleased to announce the following product improvements in our 2022.4 (Dec 2022) product release:

• Kubernetes 1.23 Support: Wallaroo version 2022.4 supports Kubernetes Version 1.23. Versions prior to the 2022.4 release do not support Kubernetes 1.23 - please check the prerequisites before installing.
• Model Serving Inference API: Wallaroo allows organizations to serve models managed in Wallaroo through an API connection from either within the Wallaroo instance’s Kubernetes internal environment, or from an external API connection. External inference endpoints must be enabled through the Wallaroo Administrative Dashboard.
• MLOps API: Wallaroo provides a set of Wallaroo MLOps API endpoints, allowing organizations to use API calls through an HTTPS connection to their Wallaroo instance. API users can perform many of the same operations as with the Wallaroo SDK including creating workspaces, administrating users, uploading models, deploying pipelines and performing inferences on deployed pipelines. For more details, see the following guides:
  • Wallaroo MLOps API Essentials Guide for user-friendly tutorials on using the Wallaroo MLOps API.
  • Wallaroo MLOps API Reference Guide for full details on the API commands.
• Public SDK Release: The Wallaroo SDK is now available for installation through the Python SDK for Wallaroo. Organizations can connect their Python programs from their locale systems or other Python environments to their Wallaroo instance with all of the same functionality and capabilities available through the Wallaroo JupyterHub and Python environment.
• SDK Install Guides for local, Google Vertex, and AzureML environments: To help new users implement the Wallaroo SDK in their preferred environments, the following guides are available with more to follow:
  • Wallaroo SDK Standard Install Guide: How to install the Wallaroo SDK in a standard Python development environment.
  • Wallaroo SDK AzureML Install Guide: How to install the Wallaroo SDK in the AzureML environment and make a connection to a Wallaroo instance through Azure Single Sign-On (SSO) authentication.
  • Wallaroo SDK Google Vertex Install Guide: How to install the Wallaroo SDK in the Google Vertex environment and and make a connection to a Wallaroo instance through Google Cloud Platform (GCP) Single Sign-On (SSO) authentication.
• Wallaroo Authentication Configuration Guides: To support users who want to use Single Sign-On (SSO) authentication to their Wallaroo instances, the following new guides demonstrate how to configure identity providers for the following platforms:
  • Wallaroo SSO for Microsoft Azure.
  • Wallaroo SSO for Google Cloud Platform.
• Assay User Interface Updates and Search Filters: The Wallaroo Dashboard interface for creating and managing assays is updated to provide more configuration options. With advance options, grouping options, bin configurations, and other details can be adjusted through the user interface. Once assays are deployed users can search assays by name or filter by Status, or sort by Creation Date or Last Run. This provides organizations the ability to create and manage assays through Wallaroo Pipeline Dashboard, the Wallaroo SDK, and the Wallaroo MLOps API.
• Helm Install: Early access instructions on installing Wallaroo using helm. These include a standard Kubernetes cloud based install, and reference documents to Helm based values.

6 - September 2022 Product Release Notes

September Release Overview

As we continue to iterate on our core capabilities, we are pleased to announce the following product improvements in our September 2022 product release :

• Global Taints and Tolerations: Wallaroo pods are installed with a default Kubernetes toleration that can be changed through the Wallaroo Administrative Dashboards. This allows Kubernetes nodes with the matching taints to prevent non-matching pods from being installed into those nodes. When combined with the Install Wallaroo to Specific Nodes guide, this allows organizations to specify what nodes to install Wallaroo pods into while keeping non-Wallaroo pods out.
• Interactive Analysis on Incoming and Model Output Data: Interactive analysis through scheduled assays provide powerful analysis tools to evaluate incoming data and data output by a model against an established baseline. Organizations can use this to track model draft and take steps to ensure accuracy of inputs to the pipeline inference process and the model’s outputs.
  • Assays can be managed through the Wallaroo SDK through methods available since June 2022.
  • The September 2022 release provides an early access user interface to create assays and view analysis reports through the Wallaroo Pipeline Dashboard.

7 - August 2022 Product Release Notes

August Release Overview

As we continue to iterate on our core capabilities, we are pleased to announce the following product improvements in our August 2022 product release:

• Air Gap Installation Support: Wallaroo Enterprise supports installation into a Kubernetes cluster that is isolated from the public internet. Organizations that need to install Wallaroo into an air gapped environment can contact their Wallaroo support representative to update their Wallaroo Enterprise license for air gap support.
• Node Containment: Organizations can install Wallaroo into specific nodes in a Kubernetes cluster.
  • This update allows organizations to tag specific nodes for Wallaroo use, while using other nodes in their cluster for other services.
• Kots 1.81 support: Wallaroo now supports Kots 1.81 for versions after the August 2022 release (code name Mustang) and beyond.
  • kots 1.70.1 is preferred for versions of Wallaroo before the August 2022 release.

8 - July 2022 Product Release Notes

July Release Overview

As we continue to iterate on our core capabilities, we are pleased to announce the following product improvements in our July 2022 product release:

• Kubernetes 1.22 Support:
  • Kubernetes version 1.22 now the preferred version for versions of Wallaroo released after the July 2022 release (code name Simca) and beyond.
• Shadow Deployments Updates:
  • Wallaroo’s shadow deployments, used for A/B Testing or parallel deployments, now include the challenger models’ outputs as part of the InferenceResult object.
  • Logging is updated to display both regular logs that include only the champion models, and shadow logs that include the shadow deployed model outputs grouped by inputs. These features allow organizations to test and compare different ML models from the same data to pick the that best fits their requirements.
  • For more details, see the blog post https://www.wallaroo.ai/blog/the-what-why-and-how-of-a/b-testing and a sample tutorial demonstrating how to deploy a shadow deployment with multiple challengers and display the results on the Wallaroo Tutorial repository.

9 - June 2022 Product Release Notes

June Release Overview

As we continue to iterate on our core capabilities, we are pleased to announce the following product improvements in our June 2022 product release:

• Model Auto-Conversion:
  • Wallaroo provides an auto-conversion method that supports sk-learn, XGBoost, and keras models. This method converts the submitted model and uploads it into the current workspace. The following tutorials demonstrate how to use the convert_model method to convert a model for use with Wallaroo and run sample inferences on the converted model:
  • sk-learn
  • keras
  • XGBoost Classification and XGBoost Regression
• Statsmodel Support:
  • Statsmodels can be imported into a Wallaroo instance as Pickle files.
• Amazon Web Services EC2 for Wallaroo Community:
  • Wallaroo Community can be installed into a Kubernetes cloud environment such as Amazon AWS, Google GCP and Microsoft Azure. This release provides support for Amazon EC2 instances for Wallaroo Community from a pre-made instance, making the installation even easier for organizations.

For more information on these and other features, see the Wallaroo Documentation Page.

For more information about this release, please contact us at deployML@wallaroo.ai .

10 - May 2022 Product Release Notes

May Release Overview

As we continue to iterate on our core capabilities, we are pleased to announce the following product improvements in our May 2022 product release:

• Pipeline User Interface Based Management
  • Available for Wallaroo Enterprise and Community customers: Pipelines are built, deployed, and undeployed from both the Wallaroo Dashboard and from the Wallaroo SDK. This allows users to click and deploy without having to run programming code.
• Single-host Linux Installation
  • Available for Wallaroo Enterprise customers Only: This feature expands Wallaroo to be installed on Kubernetes cloud environments and Kubernetes environments on single-host bare Linux and virtual machines.
• User Administration
  • For Wallaroo Enterprise customers Only: User administration is integrated into Wallaroo as a service. The capability allows enterprise administrators to set up authorized users in Wallaroo. User authentication can be configured to be either through the native Wallaroo login interface or through other Single-Sign-On services such as Google Cloud Platform, Microsoft, and GitHub. Custom Single-Sign-On solutions will be added in future versions and releases.
• DNS Configuration:
  • For Wallaroo Enterprise customers Only: Wallaroo integration with your organization’s DNS Configuration provides easier access for users to the web-based Wallaroo self-service toolkit tools (UI, API and SDK) which can be hosted in your own company domains.

For more information on these and other features, see the Wallaroo Documentation Page.

For more information about this release, please contact us at deployML@wallaroo.ai .

11 - April 2022 Product Release Notes

April release overview

As we continue to iterate on our core capabilities, we are pleased to announce the following product improvements in our April 2022 product release:

• Self-service toolkit:
  • UI Dashboard:
    • Pipeline Detail Page: The Pipeline Detail Page now has Engine Configuration details for a given pipeline deployment, and the header now shows which route the user is visiting.
    • User IDs: Dashboard now displays User Emails (or usernames as a backup), rather than their First/Last names.
• Compute engine
  • Cloud environment compute Autoscaling (Enterprise only): Supports the ability for clusters to use auto scaling node pools. This capability allows MLOps and DevOps engineers to set up dynamic compute allocation in the environment.
• Advanced Observability
  • Model Insights: Interactive baseline analysis runs the baseline in a Jupyter Notebook so that you can get a better understanding of the baseline data distribution.
  • Enhanced assay scheduling: data scientists can schedule assays on a given pipeline for model monitoring and validation checks to run at specific times with a specific frequency.

Introducing the Wallaroo Community edition:

The Wallaroo Community Edition (CE) allows users to use a free version of Wallaroo and deploy it to their clusters. The Community Edition currently has limits on the number of pipelines, number of steps in a pipeline, and number of cores available per Wallaroo instance. Wallaroo CE includes the following features:

• Simplified connectivity: Customers are given their own convenient internet address to reach their Wallaroo instance.
• Dashboard: A web based dashboard to manage users, workspaces, models, and pipelines.

Click here to learn more about Wallaroo CE.

For more information about this release, please contact us at deployML@wallaroo.ai.

12 - March 2022 Product Release Notes

March release overview

As we continue to iterate on our core capabilities, we are pleased to announce the following product improvements in our March 2022 product release:

• Compute Engine
  • Autoscaling the Wallaroo compute engine: Data Scientists can leverage the Wallaroo engine auto-scaling feature to iterate quickly on experiments. ML pipelines that Data Scientists deploy in Wallaroo will dynamically utilize allocated computational resources efficiently without any user intervention. This will ensure that available computational resources are utilized efficiently across all deployed pipelines.
• Self-Service toolkit for ML Model deployment
  • Workspaces management: Data Scientists can create and manage workspaces, in which they can invite team members to review shared models/artifacts and collaborate on model deployment, management, and monitoring efforts.

For more information about this release, please contact us at deployML@wallaroo.ai.

13 - February 2022 Product Release Notes

February release overview

As we continue to iterate on our core capabilities, we are pleased to announce the following product improvements in our February 2022 product release:

• Advanced observability:
  • Model prediction assays. Data Scientists can now create and manage validation checks that allow monitoring their ML model predictions and proactively identify data drifts.
• Self-Service toolkit for ML Model deployment:
  • Role-Based-Access-Control groups. This new security feature allows Data Scientists to manage access to ML model artifacts they own. Model artifacts can be private, public, or shared with a particular group of users.
  • Artifact management. As part of allowing users to easily retrieve their model artifacts, Data Scientists can now search model files and pipelines within a registry in the Wallaroo platform.
  • ML pipeline management. As part of this new release, we have simplified ML model deployment. In the Wallaroo platform, Data Scientists can now place their ML models into pipelines. To deploy their ML models, Data Scientists will only need to deploy the pipelines in which they placed their ML models. This capability allows Data Scientists to easily run, stop or check on the status of their model deployment activities.

For more information about this release, please contact us at deployML@wallaroo.ai.