Wallaroo Frequently Asked Questions

Wallaroo frequently asked questions

How Do I Install Wallaroo?

The Wallaroo Install Guides contain all steps on how to register with Wallaroo and receive a license, install Wallaroo into a Kubernetes cluster, and get Wallaroo running.

What Are The Prerequisites For Installing Wallaroo In A Cloud Service?

Wallaroo Community and Wallaroo Enterprise can be installed into a Kubernetes environment with the following requirements:

  • Minimum number of nodes: 4
  • Minimum Number of CPU Cores: 8
  • Minimum RAM: 16 GB
  • Kubernetes requirements:
    • Kubernetes Version
      • 1.20 is the minimum requirement
      • 1.22 is preferred version for Wallaroo versions before 2022.4.
      • 1.23:
        • Is the preferred version for Wallaroo version 2022.4.
        • Is not supported for versions of Wallaroo released before Wallaroo version 2022.4.
    • Runtime: containerd is required.

The following guide demonstrate how to set up a minimum Kubernetes environment in their cloud services and install Wallaroo:

Who Do I Contact For More Help With Wallaroo Community?

Wallaroo Community users can message community@wallaroo.ai for more assistance. Feel free to ask any questions about:

  • Installing Wallaroo
  • Uploading models and deploying pipelines
  • How to use the SDK

How Do I Set Up My Wallaroo Community User Account?

Setting up a Wallaroo Community account is fast and easy at https://portal.wallaroo.community. See the Wallaroo Install Guides for full details on registering an account and downloading your Wallaroo Community License.

How Do I Invite Collaborators And Peers To Work With After Installing Wallaroo?

Yes you can! Wallaroo Community supports up to 2 team members, while Wallaroo Enterprise has no such limitations.

See the Wallaroo User Management Guide for how to add users to your Wallaroo instance. For details on how to add users to a Wallaroo Workspace, see the Wallaroo Workspace Management Guide.

What Are The Resources I Need To Be Able To Learn About The Wallaroo Platform?

Wallaroo has provided the following resources to help you:

  • The Wallaroo 101 Guide teaches the basic concepts of how Wallaroo works, then provides a full tutorial for uploading a model and running an inference.
  • Wallaroo Tutorials are paired with the Wallaroo Tutorials Repository to provide Jupyter Notebooks that you can upload to Wallaroo along with sample models and data to see how to deploy your own ML Models.
  • The Wallaroo Essential SDK walks you the process of user, model, workspace, and pipeline management using the Wallaroo SDK.

How Fast is Wallaroo?

Not only is Wallaroo fast - its more cost efficient. Using the Aloha Tutorial as a benchmark, Wallaroo provides the following:

  # inferences per second inferences/dollar
Vertex 1.824 6,493,506
Databricks 1.856 16,000,000
SageMaker 5.008 28,169,014
Wallaroo 21.584 127,659,574

How Can I Bring My Models Into Wallaroo?

Absolutely! Wallaroo supports the ONNX ML Standard and can convert other models through our auto-conversion feature. We have created guides to help you convert some of the most popular ML model formats into ONNX. See the ONNX Conversion Tutorials for more information.

Models can be converted to native runtime models by meeting the following requirements.

Wallaroo can directly import Open Neural Network Exchange models into the Wallaroo engine. Other ML Models can be imported with the Auto-Convert Models methods.

The following models are supported natively by Wallaroo:

Wallaroo Version ONNX Version ONNX IR Version ONNX OPset Version ONNX ML Opset Version Tensorflow Version
2022.4 (December 2022) 1.12.1 8 17 3 2.9.1
After April 2022 until release 2022.4 (December 2022) 1.10.* 7 15 2 2.4
Before April 2022 1.6.* 7 13 2 2.4

For the most recent release of Wallaroo September 2022, the following native runtimes are supported:

Using different versions or settings outside of these specifications may result in inference issues and other unexpected behavior.

The following ML Model versions and Python libraries are supported by Wallaroo. When using the Wallaroo autoconversion library or working with a local version of the Wallaroo SDK, use the following versions for maximum compatibility.

Library Supported Version
Python 3.8.6 and above
onnx 1.12.0
tensorflow 2.9.1
keras 2.9.0
pytorch Latest stable version. When converting from PyTorch to onnx, verify that the onnx version matches the version above.
sk-learn aka scikit-learn 1.1.2
XGBoost 1.6.2
MLFlow 1.30.0

Can I Import My Own Notebooks Into Wallaroo?

Yes! Jupyter Hub is provided as a service in your Wallaroo instance. So feel free to import your notebooks, your ONNX mdoels, your data and get right to work.

What Is The Difference Between A Model And A Pipeline In Wallaroo?

A Model, or Machine Learning Model (ML), has been trained by data scientists to take in data and return some result. A Wallaroo Pipeline can set one or more models as steps in the pipeline. This lets you submit data to a pipeline, have that data sent to each model, then return a result.

This provides you with the power to set the order for how pipelines provide inferences in different orders, chain different models together to run comparisons, or any other combination of tasks you can come up with.

Can I Train A Model In Wallaroo?

While Wallaroo contains an entire Python library and allows you full access and control within that environment, models should be trained outside of Wallaroo, then imported into Wallaroo to be deployed and run as an object in the Wallaroo engine.

How Can I Serve My Models Up To Return Predictions?

Once you have uploaded your models to Wallaroo and deployed a pipeline that steps through how information is submitted to each model you use for the inference, you can serve your models in the following ways:

Can I A/B Test My Models In Wallaroo?

Yes! Wallaroo supports A/B testing through the following mechanisms:

  • Set up Control and Challenger models
  • Define a pipeline to split data between the two models
  • Test the Challenger model and return the results.
  • Perform a Shadow Deploy to send data to both models, but only return to the user the results from the Control model. This allows users to test Challenger models without sacrificing production deployments.

Where Can I See Pipeline Performance Metrics?

Pipeline metrics can be seen through the Wallaroo Dashboard through the following process:

  1. From the Wallaroo Dashboard, set the current workspace from the top left dropdown list.
  2. Select View Pipelines from the pipeline’s row.
  3. To view details on the pipeline, select the name of the pipeline.
  4. A list of the pipeline’s details will be displayed.
  5. Select Metrics to view the following information. From here you can select the time period to display metrics from through the drop down to display the following:
    1. Requests per second
    2. Cluster inference rate
    3. Inference latency
  6. The Audit Log and Anomaly Log are available to view further details of the pipeline’s activities.

How Are Wallaroo Enterprise And Wallaroo Community Edition Different?

Wallaroo Community is different from Wallaroo Enterprise based on the available features and restrictions.

Feature Wallaroo Community Wallaroo Enterprise
Max Number of Cores 32 Unlimited
Max Number of Users 5 Unlimited
Max Number of Deployed Pipelines 5 Unlimited
Max Steps per Pipeline 5 Unlimited
Single Sign On
Compute Auto-Scaling
Wallaroo Support Services Wallaroo Community Slack Wallaroo Enterprise Support