step_functions_mlworkflow_scikit_learn_data_processing_and_model_evaluation.ipynb failed CI

Link to the notebook:
https://github.com/aws/amazon-sagemaker-examples/blob/master/step-functions-data-science-sdk/step_functions_mlworkflow_processing/step_functions_mlworkflow_scikit_learn_data_processing_and_model_evaluation.ipynb

Error:

---------------------------------------------------------------------------
Exception encountered at "In [23]":
---------------------------------------------------------------------------
ParamValidationError Traceback (most recent call last)
in
6 )
7
----> 8 branching_workflow.create()
9
10 # Execute workflow

/opt/conda/lib/python3.7/site-packages/stepfunctions/workflow/stepfunctions.py in create(self)
203
204 try:
--> 205 self.state_machine_arn = self._create()
206 except self.client.exceptions.StateMachineAlreadyExists as e:
207 self.state_machine_arn = self._extract_state_machine_arn(e)

/opt/conda/lib/python3.7/site-packages/stepfunctions/workflow/stepfunctions.py in _create(self)
215 definition=self.definition.to_json(pretty=self.format_json),
216 roleA

[...]

Invalid length for parameter roleArn, value: 0, valid min length: 1

Start with the linked notebook and inspect cell In [23], where branching_workflow.create() fails with an empty roleArn. Reproduce the notebook failure in the CI environment, then trace how the workflow role is supplied. Done means the notebook creates the workflow successfully and its CI run passes.

Written by the indexing model from the issue text.