Yatai

As explained in Machine Learning Platform Engineering

Yatai is the BentoML component for deploying, operating, scaling, and monitoring machine-learning services on Kubernetes.

Yatai is BentoML’s Kubernetes component for deploying, operating, scaling, and monitoring machine-learning services. A typical workflow packages an application into a Bento, pushes it to a container registry, and lets Yatai build the image and deploy it. section 6.1 section 2.1.5

A practical Yatai workflow has two main stages:

• Build a Bento containing the model application and its requirements on the local system. section 6.1
• Deploy the Bento through Yatai, then use its UI to scale and monitor the resulting service. section 6.1

Income classifier API

An income classifier can be packaged as a Bento and pushed to Yatai. Yatai handles containerizing and deploying it as an API endpoint on Kubernetes, while the service exposes operational metrics such as requests per second and HTTP response codes. section 6.1 section 2.1.5

Figure 6.2 BentoML process to deploy the model as a service. We build a Bento and push it to the container registry, and Yatai builds an image and deploys it.

The deployment diagram shows how a Bento moves through the container registry before Yatai builds an image and deploys the model service. section 6.1

How Machine Learning Platform Engineering covers it

How deep it goes

Explains Bento construction and deployment plus UI operations, while deferring Kubernetes setup to an appendix.

Examples

Uses an income classifier API and a registry-based workflow for building and deploying its container image.

What it assumes

Requires BentoML in a Kubernetes cluster and a packaged Bento; does not teach cluster setup.

What you can do after

Operate/configure ML services by building a Bento and deploying it through Yatai’s Kubernetes workflow.

How it connects

Connects Bento archives, container registries, BentoML, Kubernetes, Prometheus metrics, and model deployment.

liveBookmanning.com

Read Machine Learning Platform Engineering on liveBook Get the book at manning.com

Related concepts

• Model Servingkind of · A model-serving platform is software and infrastructure that hosts trained machine-learning models, handles inference requests, and returns model outputs.
• bentoused by · A Bento is a unified package and file format that bundles a machine-learning application, its model, serving logic, dependencies, and configuration.
• BentoMLpart of · BentoML is an open-source framework for packaging, deploying, and serving machine-learning models as production services or APIs.
• computer clusteruses · A computer cluster is a group of interconnected computing nodes managed or coordinated as a system to provide shared computing capacity or services.

This explainer was generated by Manning's knowledge graph from the text of Machine Learning Platform Engineering. The section links point to the parts of the book it draws on.