MLEM now offers deployment to Kubernetes and Sagemaker with a single command.
Converting a model into MLEM model enables all kind of productionization scenarios as we have seen in Get Started.
Model extensions add support for new ML frameworks that MLEM can recognize when
you call save API method or pass a ML model to
mlem.api methods. Although MLEM can save a model from unknown ML framework
using pickle, a special extension for that framework will allow to save the
model in framework-specific format and support special features of that
framework.
Typically, model extensions will implement ModelType and ModelIO interfaces.
Some also implement DataType interface if specific data objects are needed for model to work.