create your estimator and submit it to Azure for training

azure_launch_training.py

from azureml.train.dnn import TensorFlow

datastore = ws.get_default_datastore()

script_params = {
    '--data-folder': datastore.as_mount(),
    '--batch-size': 32,
    '--learning-rate': 0.001,
    '--prefix': 'founder-classifier',
    '--steps': 1000
}

estimator = TensorFlow(source_directory=script_folder,
                 script_params=script_params,
                 compute_target=compute_target, 
                 entry_script='azure_entry_point.py',
                 pip_packages=None)

run = exp.submit(estimator)