Gradient Deployments

Gradient Deployments helps you perform effortless model serving.

Gradient Deployments is currently in Preview. If you would like to provide feedback on the deployments product, please reach out.

Overview

After Notebooks and Workflows, the third major component of the Gradient Platform is Deployments.

Creating a new deployment in the Gradient console.

Deployments is used to run container images and to serve machine learning models using a high-performance, low-latency micro-service with a RESTful API.

This allows the model to be run on new unseen data in production, also known as model inference.

Example deployment with YAML spec.

Where to start

The best place to start learning how to deploy models on Gradient is the official Gradient Deployments Tutorial:

Gradient Deployments Tutorial

Key concepts

Deployments has a number of concepts to allow different configuration.

Deployment Spec

A deployment spec is used to represent the desired state of your deployment. With the Gradient CLI you can use a YAML file to change the desired state of your deployment.

Deployment Run

A deployment can have multiple runs at the same time. Any update to the deployment spec can create a new deployment run. Once the latest deployment run is ready, the previous deployment run will scale down.