Experiment Registry

Once we finished training all the models using ml package. We can see all their performance metrics in the Experiment Registry.

To evaluate the model performances and the metrics, you need to click on the Experiments tab on the left sidebar on platform.

By using this Experiment Registry, you can evaluate the models based on their performance by looking at their metrics. Similarly, you can compare multiple models simultaneously. Once you find your best model, you can register the model and use it for Staging and Production purposes.

You can find all the experiments that you've done in the workspace. You need to choose the current experiment to view all the models that you have trained.

You can even see the versioning of the models in case you train the same model several times. It includes most of the popular evaluation metrics like accuracy score, precision score, recall score, f-1 score etc. You can use these metrics to evaluate the models and to find the best model in your experiment.

Once you've completed the training process and selected your best model by analyzing their performances, you can simply register the model so that it can be sent to either staging or production stage.

Note: Without registering the model you cannot productionize or deploy the model.

Steps to Register the model.

• So, once you find the best model, simply click on the time label of the particular model.

The models that are already registered will have different icon, you can see that in the above figure(red-colured box).

• After clicking on the time label of the best model, choose the Register Model option.

• Then we have to choose create new model and give a name for it in order to differentiate between other registered models. Your model will get registered.

Once you registered the model you can see it in the Model Registry. Now, you can use the same model for Production.