Cookbook¶

Welcome to the UbiOps cookbook!

The UbiOps cookbook is here to provide (new) users with inspiration on how to work with UbiOps. Use it to find inspiration or to discover new ways of working with the UbiOps platform.

With a (free) UbiOps account you can use the cookbook to have example applications running in your own environment in minutes.*

How does it work?¶

We have two categories in the cookbook, recipes and ready deployments. Recipes contain full walkthroughs in jupyter notebooks, and ready deployments contain ready-to-go deployment packages which illustrate how to use the deployment package for typical cases.

Ready deployments¶

steps-overview The ready deployments show how to set up your deployment package for typical use cases. You can download the deployment package, fill in the deployment creation form in the UI, and upload the deployment package. Afterwards you can make a request to the deployment to test it out.

Recipes¶

steps-overview Every recipe contains a standalone example with all the material you need to run it. They are all centered around a Jupyter Notebook. If you download the recipe folder and run the notebook it will build the example in your own UbiOps account.

The current recipes¶

Topic and link to recipe	Functionalities of UbiOps addressed
Creating a training and production pipeline with Scikit Learn in UbiOps	Deployments, pipelines
Deploying a TensorFlow model in UbiOps	Deployments
Deploying an XGBoost model in UbiOps	Deployments
Triggering a deployment/pipeline request from Azure Functions	Different forms of requests, integration
Triggering a deployment/pipeline request from Google Cloud Functions	Different forms of requests, integration
Using blobs as temporary storage	Blobs
Azure Data Factory and UbiOps pipeline interaction tutorial	Integration, pipelines
Using Azure ML services to train a model and deploy on UbiOps	Integration, deployments
Pipeline that matches, orders and visualises a list of Pokemon	Pipelines
Scheduled pipeline that classifies Amazon reviews	Request schedules, pipelines
Deploying a recommender model using Apriori in UbiOps	Deployments
Integration with MLflow model tuning tool	Deployments
Integration with snowflake cloud-based data-warehouse	Integration, deployments
Integration with YData for imbalanced datasets	Integration, deployments, blobs
Integration with Pachyderm for automated retraining	Integration, deployments
Integration with Arthur for data science monitoring	Integration, deployments

Requirements¶

To be able to use the UbiOps cookbook you need three things:

You need to have the UbiOps client library installed. This can be done via pip install or via Setuptools. For more information see our GitHub page.
You need to be able to run Jupyter Notebook. See the installation guide for more information.
You need to have a UbiOps account. You can create a free account here.

*You might need to make some space in your project by deleting deployment versions if you want to run all the examples and stay within the limits of your account.