Add a section about Local Mode to the README

README.rst

@@ -117,6 +117,78 @@ Later sections of this document explain how to use the different Estimators and
 * `Custom SageMaker Estimators and Models <#byo-docker-containers-with-sagemaker-estimators>`__
 
 
+Estimator Usage
+---------------
+
+Here is an end to end example of how to use a SageMaker Estimator.
+
+.. code:: python
+
+    from sagemaker.mxnet import MXNet
+
+    # Configure an MXNet Estimator (no training happens yet)
+    mxnet_estimator = MXNet('train.py',
+                            train_instance_type='ml.p2.xlarge',
+                            train_instance_count = 1)
+
+    # Starts a SageMaker training job and waits until completion.
+    mxnet_estimator.fit('s3://my_bucket/my_training_data/')
+
+    # Deploys the model that was generated by fit() to a SageMaker Endpoint
+    mxnet_predictor = mxnet_estimator.deploy(initial_instance_count=1, instance_type='ml.p2.xlarge')
+
+    # Serializes data and makes a prediction request to the SageMaker Endpoint
+    response = predictor.predict(data)
+
+    # Tears down the SageMaker Endpoint
+    mxnet_estimator.delete_endpoint()
+
+Local Mode
+~~~~~~~~~~
+
+The SageMaker Python SDK now supports local mode, which allows you to create TensorFlow, MXNet and BYO estimators and
+deploy to your local environment.  This is a great way to test your deep learning script before running in
+SageMaker's managed training or hosting environments.
+
+We can take the example in  `Estimator Usage <#estimator-usage>`__ , and use either ``local`` or ``local_gpu`` as the
+instance type.
+
+.. code:: python
+
+    from sagemaker.mxnet import MXNet
+
+    # Configure an MXNet Estimator (no training happens yet)
+    mxnet_estimator = MXNet('train.py',
+                            train_instance_type='local',
+                            train_instance_count=1)
+
+    # In Local Mode, fit will pull the MXNet container docker image and run it locally
+    mxnet_estimator.fit('s3://my_bucket/my_training_data/')
+
+    # Deploys the model that was generated by fit() to local endpoint in a container
+    mxnet_predictor = mxnet_estimator.deploy(initial_instance_count=1, instance_type='local')
+
+    # Serializes data and makes a prediction request to the local endpoint
+    response = predictor.predict(data)
+
+    # Tears down the endpoint container
+    mxnet_estimator.delete_endpoint()
+
+
+For detailed examples of running docker in local mode, see:
+
+- `TensorFlow local mode example notebook <https://github.com/awslabs/amazon-sagemaker-examples/blob/master/sagemaker-python-sdk/tensorflow_distributed_mnist/tensorflow_local_mode_mnist.ipynb>`__.
+- `MXNet local mode example notebook <https://github.com/awslabs/amazon-sagemaker-examples/blob/master/sagemaker-python-sdk/mxnet_gluon_mnist/mnist_with_gluon_local_mode.ipynb>`__.
+
+A few important notes:
+
+- Only one local mode endpoint can be running at a time
+- Since the data are pulled from S3 to your local environment, please ensure you have sufficient space.
+- If you run into problems, this is often due to different docker containers conflicting.  Killing these containers and re-running often solves your problems.
+- Local Mode requires docker-compose and `nvidia-docker2 <https://github.com/NVIDIA/nvidia-docker>`__ for ``local_gpu``.
+- Distributed training is not yet supported for ``local_gpu``.
+
+
 MXNet SageMaker Estimators
 --------------------------