Using Models Manager from local enviroment

Models Manager can be use for working with local develop of datascienticts. The only precondicion is some onesait platform instalation with a version of 3.1.0 or more and the direct connection from the user computer to the server.

We’ll do some examples in python code but this can be extensible to other languages compatible with MLFlow.

First thing that we need to do it’s to install some python packages that are necessary for working with Models Manager.

• MLFlow → pip install mlflow

• mlflow onesaitplatform plugin (use of the file upload tool to work with ML project files) → pip install mlflow-onesaitplatform-plugin

Second and final step is to set the tracking server of the onesait platform. This the way to connect our local code to the onesait platform server. We can do this in two different ways:

• By means of an environment variable called MLFLOW_TRACKING_URI, which we will set in the {environment}/controlpanel/modelsmanager

Open

• Through python code using the method

Open

Having a MLFlow project you can run the train of a model with the following command

mlflow run {folder project} -P {mlflowparams}

For example running the sklearn example of wine:

And watch this execution in the controlpanel and explore it:

And see the artifact of the project and the model itself

mlflow-onesaitplatform-plugin