Update files

Author: Balaji Veeramani

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()

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,

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

tests/integ/test_byo_estimator.py

@@ -88,10 +88,13 @@ def test_byo_estimator(sagemaker_session, region, cpu_instance_type, training_se
 
     with timeout_and_delete_endpoint_by_name(job_name, sagemaker_session):
         model = estimator.create_model()
-        predictor = model.deploy(1, cpu_instance_type, endpoint_name=job_name)
-        predictor.serializer = _FactorizationMachineSerializer()
-        predictor.deserializer = sagemaker.deserializers.JSONDeserializer()
-        predictor.content_type = "application/json"
+        predictor = model.deploy(
+            1,
+            cpu_instance_type,
+            endpoint_name=job_name,
+            serializer=_FactorizationMachineSerializer(),
+            deserializer=sagemaker.deserializers.JSONDeserializer(),
+        )
 
         result = predictor.predict(training_set[0][:10])
 
@@ -134,10 +137,13 @@ def test_async_byo_estimator(sagemaker_session, region, cpu_instance_type, train
             training_job_name=job_name, sagemaker_session=sagemaker_session
         )
         model = estimator.create_model()
-        predictor = model.deploy(1, cpu_instance_type, endpoint_name=endpoint_name)
-        predictor.serializer = _FactorizationMachineSerializer()
-        predictor.deserializer = sagemaker.deserializers.JSONDeserializer()
-        predictor.content_type = "application/json"
+        predictor = model.deploy(
+            1,
+            cpu_instance_type,
+            endpoint_name=endpoint_name,
+            serializer=_FactorizationMachineSerializer(),
+            deserializer=sagemaker.deserializers.JSONDeserializer(),
+        )
 
         result = predictor.predict(training_set[0][:10])
 

tests/integ/test_neo_mxnet.py

@@ -19,6 +19,7 @@
 
 from sagemaker.mxnet.estimator import MXNet
 from sagemaker.mxnet.model import MXNetModel
+from sagemaker.serializers import IdentitySerializer
 from sagemaker.utils import unique_name_from_base
 from tests.integ import DATA_DIR, TRAINING_DEFAULT_TIMEOUT_MINUTES
 from tests.integ.timeout import timeout, timeout_and_delete_endpoint_by_name
@@ -71,10 +72,14 @@ def test_attach_deploy(
             output_path=estimator.output_path,
         )
 
+        serializer = IdentitySerializer(content_type="application/vnd+python.numpy+binary")
         predictor = estimator.deploy(
-            1, cpu_instance_type, use_compiled_model=True, endpoint_name=endpoint_name
+            1,
+            cpu_instance_type,
+            serializer=serializer,
+            use_compiled_model=True,
+            endpoint_name=endpoint_name,
         )
-        predictor.content_type = "application/vnd+python.numpy+binary"
         data = numpy.zeros(shape=(1, 1, 28, 28))
         predictor.predict(data)
 
@@ -112,9 +117,11 @@ def test_deploy_model(
             job_name=unique_name_from_base("test-deploy-model-compilation-job"),
             output_path="/".join(model_data.split("/")[:-1]),
         )
-        predictor = model.deploy(1, cpu_instance_type, endpoint_name=endpoint_name)
+        serializer = IdentitySerializer(content_type="application/vnd+python.numpy+binary")
+        predictor = model.deploy(
+            1, cpu_instance_type, serializer=serializer, endpoint_name=endpoint_name
+        )
 
-        predictor.content_type = "application/vnd+python.numpy+binary"
         data = numpy.zeros(shape=(1, 1, 28, 28))
         predictor.predict(data)
 
@@ -153,8 +160,10 @@ def test_inferentia_deploy_model(
             job_name=unique_name_from_base("test-deploy-model-compilation-job"),
             output_path="/".join(model_data.split("/")[:-1]),
         )
-        predictor = model.deploy(1, inf_instance_type, endpoint_name=endpoint_name)
+        serializer = IdentitySerializer(content_type="application/vnd+python.numpy+binary")
+        predictor = model.deploy(
+            1, inf_instance_type, serializer=serializer, endpoint_name=endpoint_name
+        )
 
-        predictor.content_type = "application/vnd+python.numpy+binary"
         data = numpy.zeros(shape=(1, 1, 28, 28))
         predictor.predict(data)

tests/integ/test_tuner.py

@@ -900,10 +900,13 @@ def test_tuning_byo_estimator(sagemaker_session, cpu_instance_type):
 
     best_training_job = tuner.best_training_job()
     with timeout_and_delete_endpoint_by_name(best_training_job, sagemaker_session):
-        predictor = tuner.deploy(1, cpu_instance_type, endpoint_name=best_training_job)
-        predictor.serializer = _FactorizationMachineSerializer()
-        predictor.deserializer = JSONDeserializer()
-        predictor.content_type = "application/json"
+        predictor = tuner.deploy(
+            1,
+            cpu_instance_type,
+            endpoint_name=best_training_job,
+            serializer=_FactorizationMachineSerializer(),
+            deserializer=JSONDeserializer(),
+        )
 
         result = predictor.predict(datasets.one_p_mnist()[0][:10])
 

tests/integ/test_tuner_multi_algo.py

@@ -189,12 +189,17 @@ def _deploy_and_predict(sagemaker_session, tuner, data_set, cpu_instance_type):
                 tuner.latest_tuning_job.name, best_training_job
             )
         )
-        predictor = tuner.deploy(1, cpu_instance_type, endpoint_name=best_training_job)
-        predictor.content_type = "application/json"
+        predictor = tuner.deploy(
+            1,
+            cpu_instance_type,
+            endpoint_name=best_training_job,
+            serializer=PredictionDataSerializer(),
+            deserializer=JSONDeserializer(),
+        )
 
         print("Making prediction using the deployed model")
         data = data_set[0][:10]
-        result = _make_prediction(predictor, data)
+        result = predictor.predict(data)
 
         assert len(result["predictions"]) == len(data)
         for prediction in result["predictions"]:
@@ -214,12 +219,6 @@ def _create_training_inputs(sagemaker_session):
     return {"train": s3_train_data, "test": s3_train_data}
 
 
-def _make_prediction(predictor, data):
-    predictor.serializer = PredictionDataSerializer()
-    predictor.deserializer = JSONDeserializer()
-    return predictor.predict(data)
-
-
 class PredictionDataSerializer(BaseSerializer):
 
     CONTENT_TYPE = "application/json"

tests/unit/test_estimator.py

@@ -2020,6 +2020,8 @@ def test_generic_to_deploy_kms(create_model, sagemaker_session):
     model.deploy.assert_called_with(
         instance_type=INSTANCE_TYPE,
         initial_instance_count=INSTANCE_COUNT,
+        serializer=None,
+        deserializer=None,
         accelerator_type=None,
         endpoint_name=endpoint_name,
         tags=None,