Using managed services for deploying and using Flair models

Unlike self-serving, where most aspects of the ML life cycle need to be taken care of manually, the managed and fully managed ML services sell the idea of a complete out-of-the box ML as a service solution.

Most of these services offer guarantees about service availability (what percentage of the time the service is guaranteed to be working) and scalability (the ability to scale without having to refactor the entire infrastructure every time the user base grows). Some services also offer management of the entire ML life cycle called Machine Learning Model Operationalization (MLOps) management. But some managed services may have trouble providing support for all the features and tasks Flair is capable of solving. This applies to almost all popular ML-as-a-service solutions, with one exception – the Hugging Face Models Hub.

The Hugging Face Models Hub

Hugging Face is an NLP-oriented company with a big open source...