Enterprise Applicable to Enterprise Edition
Deploy Applicable to Deploy Edition
PrimeHub integrates Seldon as the model deployment solution. Seldon is a container-based solution and provides a common way to package different ML framework's by different programming languages as container images. We also use the Seldon Predition API to communicate.
To deploy a model, we need to provide a model image. There are two ways to prepare the model image.
- Pre-Packaged Servers: Users use a pre-packaged server as the model image. PrimeHub provides several ready-to-use pre-packaged servers. To deploy a model, a model URI is required to indicate where to download the model files.
- Package Model by Language Wrapper: Users build a model image with model files embedded. It is an advanced but flexible way to deploy a model. Seldon provides several language wrappers to package models from different languages.
In the following tutorials, we will present several different ways to deploy models.