documentation: add more info about volume_size

src/sagemaker/estimator.py

@@ -166,10 +166,44 @@ def __init__(
             instance_type (str): Type of EC2 instance to use for training,
                 for example, ``'ml.c4.xlarge'``. Required if instance_groups is
                 not set.
-            volume_size (int): Size in GB of the EBS volume to use for
-                storing input data during training (default: 30). Must be large
-                enough to store training data if File Mode is used (which is the
-                default).
+            volume_size (int): Size in GB of the storage volume to use for
+                storing input and output data during training (default: 30).
+
+                Must be large enough to store training data if File mode is
+                used, which is the default mode.
+
+                When you use an ML instance with the EBS-only storage option
+                such as ``ml.c5`` and ``ml.p2``,
+                you must define the size of the EBS
+                volume through the ``volume_size`` parameter in the estimator class.
+
+                .. note::
+
+                    When you use an ML instance with `NVMe SSD volumes
+                    <https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/ssd-instance-store.html#nvme-ssd-volumes>`_
+                    such as ``ml.p4d``, ``ml.g4dn``, and ``ml.g5``,
+                    do not include this parameter in the estimator configuration.
+                    If you use one of those ML instance types,
+                    SageMaker doesn't provision Amazon EBS General Purpose SSD
+                    (gp2) storage nor take this parameter to adjust the NVMe instance storage.
+                    Available storage is fixed to the NVMe instance storage
+                    capacity. SageMaker configures storage paths for training
+                    datasets, checkpoints, model artifacts, and outputs to use the
+                    entire capacity of the instance storage.
+
+                    Note that if you include this parameter and specify a number that
+                    exceeds the size of the NVMe volume attached to the instance type,
+                    SageMaker returns an ``Invalid VolumeSizeInGB`` error.
+
+                To look up instance types and their instance storage types
+                and volumes, see `Amazon EC2 Instance Types
+                <http://aws.amazon.com/ec2/instance-types/>`_.
+
+                To find the default local paths defined by the SageMaker
+                training platform, see `Amazon SageMaker Training Storage
+                Folders for Training Datasets, Checkpoints, Model Artifacts,
+                and Outputs
+                <https://docs.aws.amazon.com/sagemaker/latest/dg/model-train-storage.html>`_.
             volume_kms_key (str): Optional. KMS key ID for encrypting EBS
                 volume attached to the training instance (default: None).
             max_run (int): Timeout in seconds for training (default: 24 *
@@ -2232,12 +2266,46 @@ def __init__(
             instance_count (int): Number of Amazon EC2 instances to use
                 for training. Required if instance_groups is not set.
             instance_type (str): Type of EC2 instance to use for training,
-                for example, 'ml.c4.xlarge'. Required if instance_groups is
+                for example, ``'ml.c4.xlarge'``. Required if instance_groups is
                 not set.
-            volume_size (int): Size in GB of the EBS volume to use for
-                storing input data during training (default: 30). Must be large
-                enough to store training data if File Mode is used (which is the
-                default).
+            volume_size (int): Size in GB of the storage volume to use for
+                storing input and output data during training (default: 30).
+
+                Must be large enough to store training data if File mode is
+                used, which is the default mode.
+
+                When you use an ML instance with the EBS-only storage option
+                such as ``ml.c5`` and ``ml.p2``,
+                you must define the size of the EBS
+                volume through the ``volume_size`` parameter in the estimator class.
+
+                .. note::
+
+                    When you use an ML instance with `NVMe SSD volumes
+                    <https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/ssd-instance-store.html#nvme-ssd-volumes>`_
+                    such as ``ml.p4d``, ``ml.g4dn``, and ``ml.g5``,
+                    do not include this parameter in the estimator configuration.
+                    If you use one of those ML instance types,
+                    SageMaker doesn't provision Amazon EBS General Purpose SSD
+                    (gp2) storage nor take this parameter to adjust the NVMe instance storage.
+                    Available storage is fixed to the NVMe instance storage
+                    capacity. SageMaker configures storage paths for training
+                    datasets, checkpoints, model artifacts, and outputs to use the
+                    entire capacity of the instance storage.
+
+                    Note that if you include this parameter and specify a number that
+                    exceeds the size of the NVMe volume attached to the instance type,
+                    SageMaker returns an ``Invalid VolumeSizeInGB`` error.
+
+                To look up instance types and their instance storage types
+                and volumes, see `Amazon EC2 Instance Types
+                <http://aws.amazon.com/ec2/instance-types/>`_.
+
+                To find the default local paths defined by the SageMaker
+                training platform, see `Amazon SageMaker Training Storage
+                Folders for Training Datasets, Checkpoints, Model Artifacts,
+                and Outputs
+                <https://docs.aws.amazon.com/sagemaker/latest/dg/model-train-storage.html>`_.
             volume_kms_key (str): Optional. KMS key ID for encrypting EBS
                 volume attached to the training instance (default: None).
             max_run (int): Timeout in seconds for training (default: 24 *