Deploying Models to Production
After creating an API that contains your machine learning model, it has to be hosted in a production environment. There are several ways to do this, such as the following, for example:
- By copying the API to a (virtual) server
- By containerizing the API and deploying the container to a cluster
- By installing the API in a serverless framework such as Amazon AWS Lambda or Microsoft Azure Functions
We’ll focus on the practice that is very common nowadays and still gaining popularity: containerizing the API and model.
Docker
AI applications usually work with large datasets. With “big data” comes the requirement for scalability. This means that models in production should scale in line with the data. One way to scale your software services is to distribute them in containers. A container is a small unit of computational power, similar to a virtual machine. There are many other advantages when containerizing your...
