breaking: Add parameters to deploy and remove parameters from create_model (#1783)

Author: bveeramani

doc/frameworks/tensorflow/upgrade_from_legacy.rst

@@ -248,7 +248,6 @@ For example, if you want to use JSON serialization and deserialization:
     from sagemaker.deserializers import JSONDeserializer
     from sagemaker.serializers import JSONSerializer
 
-    predictor.serializer = JSONSerializer()
-    predictor.deserializer = JSONDeserializer()
+    predictor = model.deploy(..., serializer=JSONSerializer(), deserializer=JSONDeserializer())
 
     predictor.predict(data)

doc/frameworks/xgboost/using_xgboost.rst

@@ -192,12 +192,14 @@ inference against your model.
 
 .. code::
 
+    serializer = StringSerializer()
+    serializer.CONTENT_TYPE = "text/libsvm"
+
     predictor = estimator.deploy(
         initial_instance_count=1,
-        instance_type="ml.m5.xlarge"
+        instance_type="ml.m5.xlarge",
+        serializer=serializer
     )
-    predictor.serializer = str
-    predictor.content_type = "text/libsvm"
 
     with open("abalone") as f:
         payload = f.read()

doc/v2.rst

@@ -81,6 +81,13 @@ Please instantiate the objects instead.
 The ``update_endpoint`` argument in ``deploy()`` methods for estimators and models has been deprecated.
 Please use :func:`sagemaker.predictor.Predictor.update_endpoint` instead.
 
+``serializer`` and ``deserializer`` in ``create_model()``
+---------------------------------------------------------
+
+The ``serializer`` and ``deserializer`` arguments in
+:func:`sagemaker.estimator.Estimator.create_model` have been deprecated. Please
+specify serializers and deserializers in ``deploy()`` methods instead.
+
 ``content_type`` and ``accept`` in the Predictor Constructor
 ------------------------------------------------------------
 

src/sagemaker/automl/automl.py

@@ -337,6 +337,8 @@ def deploy(
         self,
         initial_instance_count,
         instance_type,
+        serializer=None,
+        deserializer=None,
         candidate=None,
         sagemaker_session=None,
         name=None,
@@ -356,6 +358,16 @@ def deploy(
                 in the ``Endpoint`` created from this ``Model``.
             instance_type (str): The EC2 instance type to deploy this Model to.
                 For example, 'ml.p2.xlarge'.
+            serializer (:class:`~sagemaker.serializers.BaseSerializer`): A
+                serializer object, used to encode data for an inference endpoint
+                (default: None). If ``serializer`` is not None, then
+                ``serializer`` will override the default serializer. The
+                default serializer is set by the ``predictor_cls``.
+            deserializer (:class:`~sagemaker.deserializers.BaseDeserializer`): A
+                deserializer object, used to decode data from an inference
+                endpoint (default: None). If ``deserializer`` is not None, then
+                ``deserializer`` will override the default deserializer. The
+                default deserializer is set by the ``predictor_cls``.
             candidate (CandidateEstimator or dict): a CandidateEstimator used for deploying
                 to a SageMaker Inference Pipeline. If None, the best candidate will
                 be used. If the candidate input is a dict, a CandidateEstimator will be
@@ -405,6 +417,8 @@ def deploy(
         return model.deploy(
             initial_instance_count=initial_instance_count,
             instance_type=instance_type,
+            serializer=serializer,
+            deserializer=deserializer,
             endpoint_name=endpoint_name,
             tags=tags,
             wait=wait,

src/sagemaker/estimator.py

@@ -29,9 +29,7 @@
 from sagemaker.debugger import DebuggerHookConfig
 from sagemaker.debugger import TensorBoardOutputConfig  # noqa: F401 # pylint: disable=unused-import
 from sagemaker.debugger import get_rule_container_image_uri
-from sagemaker.deserializers import BytesDeserializer
 from sagemaker.s3 import S3Uploader, parse_s3_url
-from sagemaker.serializers import IdentitySerializer
 
 from sagemaker.fw_utils import (
     tar_and_upload_dir,
@@ -670,6 +668,8 @@ def deploy(
         self,
         initial_instance_count,
         instance_type,
+        serializer=None,
+        deserializer=None,
         accelerator_type=None,
         endpoint_name=None,
         use_compiled_model=False,
@@ -691,6 +691,16 @@ def deploy(
                 deploy to an endpoint for prediction.
             instance_type (str): Type of EC2 instance to deploy to an endpoint
                 for prediction, for example, 'ml.c4.xlarge'.
+            serializer (:class:`~sagemaker.serializers.BaseSerializer`): A
+                serializer object, used to encode data for an inference endpoint
+                (default: None). If ``serializer`` is not None, then
+                ``serializer`` will override the default serializer. The
+                default serializer is set by the ``predictor_cls``.
+            deserializer (:class:`~sagemaker.deserializers.BaseDeserializer`): A
+                deserializer object, used to decode data from an inference
+                endpoint (default: None). If ``deserializer`` is not None, then
+                ``deserializer`` will override the default deserializer. The
+                default deserializer is set by the ``predictor_cls``.
             accelerator_type (str): Type of Elastic Inference accelerator to
                 attach to an endpoint for model loading and inference, for
                 example, 'ml.eia1.medium'. If not specified, no Elastic
@@ -753,6 +763,8 @@ def deploy(
         return model.deploy(
             instance_type=instance_type,
             initial_instance_count=initial_instance_count,
+            serializer=serializer,
+            deserializer=deserializer,
             accelerator_type=accelerator_type,
             endpoint_name=endpoint_name,
             tags=tags or self.tags,
@@ -1367,8 +1379,6 @@ def create_model(
         role=None,
         image_uri=None,
         predictor_cls=None,
-        serializer=IdentitySerializer(),
-        deserializer=BytesDeserializer(),
         vpc_config_override=vpc_utils.VPC_CONFIG_DEFAULT,
         **kwargs
     ):
@@ -1386,12 +1396,6 @@ def create_model(
                 Defaults to the image used for training.
             predictor_cls (Predictor): The predictor class to use when
                 deploying the model.
-            serializer (:class:`~sagemaker.serializers.BaseSerializer`): A
-                serializer object, used to encode data for an inference endpoint
-                (default: :class:`~sagemaker.serializers.IdentitySerializer`).
-            deserializer (:class:`~sagemaker.deserializers.BaseDeserializer`): A
-                deserializer object, used to decode data from an inference
-                endpoint (default: :class:`~sagemaker.deserializers.BytesDeserializer`).
             vpc_config_override (dict[str, list[str]]): Optional override for VpcConfig set on
                 the model.
                 Default: use subnets and security groups from this Estimator.
@@ -1410,7 +1414,7 @@ def create_model(
         if predictor_cls is None:
 
             def predict_wrapper(endpoint, session):
-                return Predictor(endpoint, session, serializer, deserializer)
+                return Predictor(endpoint, session)
 
             predictor_cls = predict_wrapper
 

src/sagemaker/model.py

@@ -407,6 +407,8 @@ def deploy(
         self,
         initial_instance_count,
         instance_type,
+        serializer=None,
+        deserializer=None,
         accelerator_type=None,
         endpoint_name=None,
         tags=None,
@@ -433,6 +435,16 @@ def deploy(
                 in the ``Endpoint`` created from this ``Model``.
             instance_type (str): The EC2 instance type to deploy this Model to.
                 For example, 'ml.p2.xlarge', or 'local' for local mode.
+            serializer (:class:`~sagemaker.serializers.BaseSerializer`): A
+                serializer object, used to encode data for an inference endpoint
+                (default: None). If ``serializer`` is not None, then
+                ``serializer`` will override the default serializer. The
+                default serializer is set by the ``predictor_cls``.
+            deserializer (:class:`~sagemaker.deserializers.BaseDeserializer`): A
+                deserializer object, used to decode data from an inference
+                endpoint (default: None). If ``deserializer`` is not None, then
+                ``deserializer`` will override the default deserializer. The
+                default deserializer is set by the ``predictor_cls``.
             accelerator_type (str): Type of Elastic Inference accelerator to
                 deploy this model for model loading and inference, for example,
                 'ml.eia1.medium'. If not specified, no Elastic Inference
@@ -499,7 +511,12 @@ def deploy(
         )
 
         if self.predictor_cls:
-            return self.predictor_cls(self.endpoint_name, self.sagemaker_session)
+            predictor = self.predictor_cls(self.endpoint_name, self.sagemaker_session)
+            if serializer:
+                predictor.serializer = serializer
+            if deserializer:
+                predictor.deserializer = deserializer
+            return predictor
         return None
 
     def transformer(

src/sagemaker/multidatamodel.py

@@ -143,6 +143,8 @@ def deploy(
         self,
         initial_instance_count,
         instance_type,
+        serializer=None,
+        deserializer=None,
         accelerator_type=None,
         endpoint_name=None,
         tags=None,
@@ -171,6 +173,16 @@ def deploy(
                 in the ``Endpoint`` created from this ``Model``.
             instance_type (str): The EC2 instance type to deploy this Model to.
                 For example, 'ml.p2.xlarge', or 'local' for local mode.
+            serializer (:class:`~sagemaker.serializers.BaseSerializer`): A
+                serializer object, used to encode data for an inference endpoint
+                (default: None). If ``serializer`` is not None, then
+                ``serializer`` will override the default serializer. The
+                default serializer is set by the ``predictor_cls``.
+            deserializer (:class:`~sagemaker.deserializers.BaseDeserializer`): A
+                deserializer object, used to decode data from an inference
+                endpoint (default: None). If ``deserializer`` is not None, then
+                ``deserializer`` will override the default deserializer. The
+                default deserializer is set by the ``predictor_cls``.
             accelerator_type (str): Type of Elastic Inference accelerator to
                 deploy this model for model loading and inference, for example,
                 'ml.eia1.medium'. If not specified, no Elastic Inference
@@ -201,12 +213,12 @@ def deploy(
             enable_network_isolation = self.model.enable_network_isolation()
             role = self.model.role
             vpc_config = self.model.vpc_config
-            predictor = self.model.predictor_cls
+            predictor_cls = self.model.predictor_cls
         else:
             enable_network_isolation = self.enable_network_isolation()
             role = self.role
             vpc_config = self.vpc_config
-            predictor = self.predictor_cls
+            predictor_cls = self.predictor_cls
 
         if role is None:
             raise ValueError("Role can not be null for deploying a model")
@@ -245,8 +257,13 @@ def deploy(
             data_capture_config_dict=data_capture_config_dict,
         )
 
-        if predictor:
-            return predictor(self.endpoint_name, self.sagemaker_session)
+        if predictor_cls:
+            predictor = predictor_cls(self.endpoint_name, self.sagemaker_session)
+            if serializer:
+                predictor.serializer = serializer
+            if deserializer:
+                predictor.deserializer = deserializer
+            return predictor
         return None
 
     def add_model(self, model_data_source, model_data_path=None):

src/sagemaker/pipeline.py

@@ -85,6 +85,8 @@ def deploy(
         self,
         initial_instance_count,
         instance_type,
+        serializer=None,
+        deserializer=None,
         endpoint_name=None,
         tags=None,
         wait=True,
@@ -110,6 +112,16 @@ def deploy(
                 in the ``Endpoint`` created from this ``Model``.
             instance_type (str): The EC2 instance type to deploy this Model to.
                 For example, 'ml.p2.xlarge'.
+            serializer (:class:`~sagemaker.serializers.BaseSerializer`): A
+                serializer object, used to encode data for an inference endpoint
+                (default: None). If ``serializer`` is not None, then
+                ``serializer`` will override the default serializer. The
+                default serializer is set by the ``predictor_cls``.
+            deserializer (:class:`~sagemaker.deserializers.BaseDeserializer`): A
+                deserializer object, used to decode data from an inference
+                endpoint (default: None). If ``deserializer`` is not None, then
+                ``deserializer`` will override the default deserializer. The
+                default deserializer is set by the ``predictor_cls``.
             endpoint_name (str): The name of the endpoint to create (default:
                 None). If not specified, a unique endpoint name will be created.
             tags (List[dict[str, str]]): The list of tags to attach to this
@@ -171,7 +183,12 @@ def deploy(
             )
 
         if self.predictor_cls:
-            return self.predictor_cls(self.endpoint_name, self.sagemaker_session)
+            predictor = self.predictor_cls(self.endpoint_name, self.sagemaker_session)
+            if serializer:
+                predictor.serializer = serializer
+            if deserializer:
+                predictor.deserializer = deserializer
+            return predictor
         return None
 
     def _create_sagemaker_pipeline_model(self, instance_type):

src/sagemaker/predictor.py

@@ -62,8 +62,6 @@ def __init__(
         self.sagemaker_session = sagemaker_session or Session()
         self.serializer = serializer
         self.deserializer = deserializer
-        self.content_type = serializer.CONTENT_TYPE
-        self.accept = deserializer.ACCEPT
         self._endpoint_config_name = self._get_endpoint_config_name()
         self._model_names = self._get_model_names()
 
@@ -380,3 +378,13 @@ def _get_model_names(self):
         )
         production_variants = endpoint_config["ProductionVariants"]
         return [d["ModelName"] for d in production_variants]
+
+    @property
+    def content_type(self):
+        """The MIME type of the data sent to the inference endpoint."""
+        return self.serializer.CONTENT_TYPE
+
+    @property
+    def accept(self):
+        """The content type(s) that are expected from the inference endpoint."""
+        return self.deserializer.ACCEPT

src/sagemaker/tensorflow/model.py

@@ -196,6 +196,8 @@ def deploy(
         self,
         initial_instance_count,
         instance_type,
+        serializer=None,
+        deserializer=None,
         accelerator_type=None,
         endpoint_name=None,
         tags=None,
@@ -211,6 +213,8 @@ def deploy(
         return super(TensorFlowModel, self).deploy(
             initial_instance_count=initial_instance_count,
             instance_type=instance_type,
+            serializer=serializer,
+            deserializer=deserializer,
             accelerator_type=accelerator_type,
             endpoint_name=endpoint_name,
             tags=tags,

src/sagemaker/tuner.py

@@ -674,6 +674,8 @@ def deploy(
         self,
         initial_instance_count,
         instance_type,
+        serializer=None,
+        deserializer=None,
         accelerator_type=None,
         endpoint_name=None,
         wait=True,
@@ -693,6 +695,16 @@ def deploy(
                 deploy to an endpoint for prediction.
             instance_type (str): Type of EC2 instance to deploy to an endpoint
                 for prediction, for example, 'ml.c4.xlarge'.
+            serializer (:class:`~sagemaker.serializers.BaseSerializer`): A
+                serializer object, used to encode data for an inference endpoint
+                (default: None). If ``serializer`` is not None, then
+                ``serializer`` will override the default serializer. The
+                default serializer is set by the ``predictor_cls``.
+            deserializer (:class:`~sagemaker.deserializers.BaseDeserializer`): A
+                deserializer object, used to decode data from an inference
+                endpoint (default: None). If ``deserializer`` is not None, then
+                ``deserializer`` will override the default deserializer. The
+                default deserializer is set by the ``predictor_cls``.
             accelerator_type (str): Type of Elastic Inference accelerator to
                 attach to an endpoint for model loading and inference, for
                 example, 'ml.eia1.medium'. If not specified, no Elastic
@@ -727,6 +739,8 @@ def deploy(
         return best_estimator.deploy(
             initial_instance_count=initial_instance_count,
             instance_type=instance_type,
+            serializer=serializer,
+            deserializer=deserializer,
             accelerator_type=accelerator_type,
             endpoint_name=endpoint_name or best_training_job["TrainingJobName"],
             wait=wait,