none sub flavor

To save or log MATLAB/Simulink code and data into a MATLAB flavor MLflow model use mlflow.matlab.save_model or mlflow.matlab.log_model with the MATLABFiles option. The option accepts a string array of separate files as well as directories. For example to save a model:

% Save two specific files and an entire directory into a MATLAB flavor MLflow model.
% When saving also specify Path to indicate where to save the model
mlflow.matlab.save_model(MATLABFiles=["someFunction.m","/some/path/file.m","/some/directory"], ...
    Path="/work/models/myModel")

Or to log a model:

% In order to log a model in a specific run in a specific experiment, first set an experiment
py.mlflow.set_experiment("My MATLAB Experiment")
% And start a run
py.mlflow.start_run()
% Then log the model, when logging a model, Path is not required
mlflow.matlab.log_model(MATLABFiles=["someFunction.m","/some/path/file.m","/some/directory"])
% End the run
py.mlflow.end_run()

In order to be able to use a saved model in MATLAB again, use mlflow.matlab.load_model to first (if needed) download or copy the model locally, and then obtain the path of the local model. Then for example either cd to that path or use addpath/genpath to add this path to the MATLAB path.

% Load a model
p = mlflow.model.load_model("/work/models/myModel")
% Add the directory structure to the MATLAB path such that the code can be run in MATLAB
addpath(genpath(p))

With a local model this will typically just return the local location where the MATLAB code and data is already located. For a model on a remote tracking server, the model will be downloaded to a temporary local location and then this location is returned. To download a remote model to a specific local location (or to make a copy of a local model), specify this desired location as second input:

% Download model to a specific location
p = mlflow.model.load_model("models:/someModel/1","/work/models/someModel")
% In this example go to that directory instead of adding it to the MATLAB path
cd(p)