MLflow models

Description:

• Standard format for packaging machine learning models that can be used in a variety of downstream tools

Storage format:

• MLFlow flavor
  • All of the flavors that a particular model supports are defined in its MLmodel file in YAML format
• If you wish to serve a model from inside a Docker container (or to query it from another machine)
  • you need to change the network address to 0.0.0.0 using the -h argument.

    mlflow models serve -h 0.0.0.0 -m my_model

Model signature and input examples:

• Model inference param:
  • description of params used for model inference
• Model signature