Lab421.md

Create an automl model against existing Synapse data

Make sure you complete Lab 420 first

• Find the dataset you created in Lab 420 called nyc_taxi_prepped under Data|Workspace in your Spark database.
• Right click that dataset
• Enrich with new model

This is automl.

Read the instructions and follow along.

• The target column this time is fare_amount. (We'll do something different this time)
• We will use Regression
• Choose 0.5 hours
• ONNX = Enabled
• Open in notebook so we can see the code

When we execute the code the session will timeout before we are able to get the best model and register it (this is the last cell in the notebook). That's ok, we can still do this.

Create a new cell that looks like this at the bottom. You will need to change the variables:

import azureml.core

from azureml.core import Experiment, Workspace, Dataset, Datastore
from azureml.train.automl import AutoMLConfig
from azureml.train.automl.run import AutoMLRun
from azureml.data.dataset_factory import TabularDatasetFactory

#changeme
subscription_id = "52061d21-01dd-4f9e-aca9-60fff4d67ee2"
resource_group = "MLOpsWorkshop"
workspace_name = "mlops"
experiment_name = "nyc_taxi_prepped-automl"
runid = "AutoML_cf62735b-06a8-4ea2-bb19-92f8b43c3d18"  #from amls portal

ws = Workspace(subscription_id = subscription_id, resource_group = resource_group, workspace_name = workspace_name)
experiment = Experiment(ws, experiment_name)
run = AutoMLRun(experiment=experiment,run_id=runid)

Now run should be set properly to register the model in the last cell of the auto-generated notebook.