feature: all predictors support serializer/deserializer overrides (#1997)

* feat: framework predictor de/serial override args

All framework *Predictor classes accept constructor arguments to
override the default `serializer` & `deserializer` logic (like
TensorFlowPredictor did already).

* feat: Amz algo predictor de/serial override args

All Amazon algorithm *Predictor classes accept constructor arguments
to override the default `serializer` & `deserializer` logic, in case
users need to specify alternative inference formats.

* change: add unit tests for de/ser overrides

Co-authored-by: Ajay Karpur <[email protected]>

Author: athewsey

src/sagemaker/amazon/factorization_machines.py

@@ -277,13 +277,20 @@ class FactorizationMachinesPredictor(Predictor):
     to fit the model this Predictor performs inference on.
 
     :meth:`predict()` returns a list of
-    :class:`~sagemaker.amazon.record_pb2.Record` objects, one for each row in
+    :class:`~sagemaker.amazon.record_pb2.Record` objects (assuming the default
+    recordio-protobuf ``deserializer`` is used), one for each row in
     the input ``ndarray``. The prediction is stored in the ``"score"`` key of
     the ``Record.label`` field. Please refer to the formats details described:
     https://docs.aws.amazon.com/sagemaker/latest/dg/fm-in-formats.html
     """
 
-    def __init__(self, endpoint_name, sagemaker_session=None):
+    def __init__(
+        self,
+        endpoint_name,
+        sagemaker_session=None,
+        serializer=RecordSerializer(),
+        deserializer=RecordDeserializer(),
+    ):
         """
         Args:
             endpoint_name (str): Name of the Amazon SageMaker endpoint to which
@@ -292,12 +299,16 @@ def __init__(self, endpoint_name, sagemaker_session=None):
                 object, used for SageMaker interactions (default: None). If not
                 specified, one is created using the default AWS configuration
                 chain.
+            serializer (sagemaker.serializers.BaseSerializer): Optional. Default
+                serializes input data to x-recordio-protobuf format.
+            deserializer (sagemaker.deserializers.BaseDeserializer): Optional.
+                Default parses responses from x-recordio-protobuf format.
         """
         super(FactorizationMachinesPredictor, self).__init__(
             endpoint_name,
             sagemaker_session,
-            serializer=RecordSerializer(),
-            deserializer=RecordDeserializer(),
+            serializer=serializer,
+            deserializer=deserializer,
         )
 
 

src/sagemaker/amazon/ipinsights.py

@@ -191,7 +191,13 @@ class IPInsightsPredictor(Predictor):
     second column should contain the IPv4 address in dot notation.
     """
 
-    def __init__(self, endpoint_name, sagemaker_session=None):
+    def __init__(
+        self,
+        endpoint_name,
+        sagemaker_session=None,
+        serializer=CSVSerializer(),
+        deserializer=JSONDeserializer(),
+    ):
         """
         Args:
             endpoint_name (str): Name of the Amazon SageMaker endpoint to which
@@ -200,12 +206,16 @@ def __init__(self, endpoint_name, sagemaker_session=None):
                 object, used for SageMaker interactions (default: None). If not
                 specified, one is created using the default AWS configuration
                 chain.
+            serializer (sagemaker.serializers.BaseSerializer): Optional. Default
+                serializes input data to text/csv.
+            deserializer (callable): Optional. Default parses JSON responses
+                using ``json.load(...)``.
         """
         super(IPInsightsPredictor, self).__init__(
             endpoint_name,
             sagemaker_session,
-            serializer=CSVSerializer(),
-            deserializer=JSONDeserializer(),
+            serializer=serializer,
+            deserializer=deserializer,
         )
 
 

src/sagemaker/amazon/kmeans.py

@@ -210,12 +210,19 @@ class KMeansPredictor(Predictor):
     to fit the model this Predictor performs inference on.
 
     ``predict()`` returns a list of
-    :class:`~sagemaker.amazon.record_pb2.Record` objects, one for each row in
+    :class:`~sagemaker.amazon.record_pb2.Record` objects (assuming the default
+    recordio-protobuf ``deserializer`` is used), one for each row in
     the input ``ndarray``. The nearest cluster is stored in the
     ``closest_cluster`` key of the ``Record.label`` field.
     """
 
-    def __init__(self, endpoint_name, sagemaker_session=None):
+    def __init__(
+        self,
+        endpoint_name,
+        sagemaker_session=None,
+        serializer=RecordSerializer(),
+        deserializer=RecordDeserializer(),
+    ):
         """
         Args:
             endpoint_name (str): Name of the Amazon SageMaker endpoint to which
@@ -224,12 +231,16 @@ def __init__(self, endpoint_name, sagemaker_session=None):
                 object, used for SageMaker interactions (default: None). If not
                 specified, one is created using the default AWS configuration
                 chain.
+            serializer (sagemaker.serializers.BaseSerializer): Optional. Default
+                serializes input data to x-recordio-protobuf format.
+            deserializer (sagemaker.deserializers.BaseDeserializer): Optional.
+                Default parses responses from x-recordio-protobuf format.
         """
         super(KMeansPredictor, self).__init__(
             endpoint_name,
             sagemaker_session,
-            serializer=RecordSerializer(),
-            deserializer=RecordDeserializer(),
+            serializer=serializer,
+            deserializer=deserializer,
         )
 
 

src/sagemaker/amazon/knn.py

@@ -199,12 +199,19 @@ class KNNPredictor(Predictor):
     to fit the model this Predictor performs inference on.
 
     :func:`predict` returns a list of
-    :class:`~sagemaker.amazon.record_pb2.Record` objects, one for each row in
+    :class:`~sagemaker.amazon.record_pb2.Record` objects (assuming the default
+    recordio-protobuf ``deserializer`` is used), one for each row in
     the input ``ndarray``. The prediction is stored in the ``"predicted_label"``
     key of the ``Record.label`` field.
     """
 
-    def __init__(self, endpoint_name, sagemaker_session=None):
+    def __init__(
+        self,
+        endpoint_name,
+        sagemaker_session=None,
+        serializer=RecordSerializer(),
+        deserializer=RecordDeserializer(),
+    ):
         """
         Args:
             endpoint_name (str): Name of the Amazon SageMaker endpoint to which
@@ -213,12 +220,16 @@ def __init__(self, endpoint_name, sagemaker_session=None):
                 object, used for SageMaker interactions (default: None). If not
                 specified, one is created using the default AWS configuration
                 chain.
+            serializer (sagemaker.serializers.BaseSerializer): Optional. Default
+                serializes input data to x-recordio-protobuf format.
+            deserializer (sagemaker.deserializers.BaseDeserializer): Optional.
+                Default parses responses from x-recordio-protobuf format.
         """
         super(KNNPredictor, self).__init__(
             endpoint_name,
             sagemaker_session,
-            serializer=RecordSerializer(),
-            deserializer=RecordDeserializer(),
+            serializer=serializer,
+            deserializer=deserializer,
         )
 
 

src/sagemaker/amazon/lda.py

@@ -183,12 +183,19 @@ class LDAPredictor(Predictor):
     to fit the model this Predictor performs inference on.
 
     :meth:`predict()` returns a list of
-    :class:`~sagemaker.amazon.record_pb2.Record` objects, one for each row in
+    :class:`~sagemaker.amazon.record_pb2.Record` objects (assuming the default
+    recordio-protobuf ``deserializer`` is used), one for each row in
     the input ``ndarray``. The lower dimension vector result is stored in the
     ``projection`` key of the ``Record.label`` field.
     """
 
-    def __init__(self, endpoint_name, sagemaker_session=None):
+    def __init__(
+        self,
+        endpoint_name,
+        sagemaker_session=None,
+        serializer=RecordSerializer(),
+        deserializer=RecordDeserializer(),
+    ):
         """
         Args:
             endpoint_name (str): Name of the Amazon SageMaker endpoint to which
@@ -197,12 +204,16 @@ def __init__(self, endpoint_name, sagemaker_session=None):
                 object, used for SageMaker interactions (default: None). If not
                 specified, one is created using the default AWS configuration
                 chain.
+            serializer (sagemaker.serializers.BaseSerializer): Optional. Default
+                serializes input data to x-recordio-protobuf format.
+            deserializer (sagemaker.deserializers.BaseDeserializer): Optional.
+                Default parses responses from x-recordio-protobuf format.
         """
         super(LDAPredictor, self).__init__(
             endpoint_name,
             sagemaker_session,
-            serializer=RecordSerializer(),
-            deserializer=RecordDeserializer(),
+            serializer=serializer,
+            deserializer=deserializer,
         )
 
 

src/sagemaker/amazon/linear_learner.py

@@ -444,12 +444,19 @@ class LinearLearnerPredictor(Predictor):
     to fit the model this Predictor performs inference on.
 
     :func:`predict` returns a list of
-    :class:`~sagemaker.amazon.record_pb2.Record` objects, one for each row in
+    :class:`~sagemaker.amazon.record_pb2.Record` objects (assuming the default
+    recordio-protobuf ``deserializer`` is used), one for each row in
     the input ``ndarray``. The prediction is stored in the ``"predicted_label"``
     key of the ``Record.label`` field.
     """
 
-    def __init__(self, endpoint_name, sagemaker_session=None):
+    def __init__(
+        self,
+        endpoint_name,
+        sagemaker_session=None,
+        serializer=RecordSerializer(),
+        deserializer=RecordDeserializer(),
+    ):
         """
         Args:
             endpoint_name (str): Name of the Amazon SageMaker endpoint to which
@@ -458,12 +465,16 @@ def __init__(self, endpoint_name, sagemaker_session=None):
                 object, used for SageMaker interactions (default: None). If not
                 specified, one is created using the default AWS configuration
                 chain.
+            serializer (sagemaker.serializers.BaseSerializer): Optional. Default
+                serializes input data to x-recordio-protobuf format.
+            deserializer (sagemaker.deserializers.BaseDeserializer): Optional.
+                Default parses responses from x-recordio-protobuf format.
         """
         super(LinearLearnerPredictor, self).__init__(
             endpoint_name,
             sagemaker_session,
-            serializer=RecordSerializer(),
-            deserializer=RecordDeserializer(),
+            serializer=serializer,
+            deserializer=deserializer,
         )
 
 

src/sagemaker/amazon/ntm.py

@@ -212,12 +212,19 @@ class NTMPredictor(Predictor):
     to fit the model this Predictor performs inference on.
 
     :meth:`predict()` returns a list of
-    :class:`~sagemaker.amazon.record_pb2.Record` objects, one for each row in
+    :class:`~sagemaker.amazon.record_pb2.Record` objects (assuming the default
+    recordio-protobuf ``deserializer`` is used), one for each row in
     the input ``ndarray``. The lower dimension vector result is stored in the
     ``projection`` key of the ``Record.label`` field.
     """
 
-    def __init__(self, endpoint_name, sagemaker_session=None):
+    def __init__(
+        self,
+        endpoint_name,
+        sagemaker_session=None,
+        serializer=RecordSerializer(),
+        deserializer=RecordDeserializer(),
+    ):
         """
         Args:
             endpoint_name (str): Name of the Amazon SageMaker endpoint to which
@@ -226,12 +233,16 @@ def __init__(self, endpoint_name, sagemaker_session=None):
                 object, used for SageMaker interactions (default: None). If not
                 specified, one is created using the default AWS configuration
                 chain.
+            serializer (sagemaker.serializers.BaseSerializer): Optional. Default
+                serializes input data to x-recordio-protobuf format.
+            deserializer (sagemaker.deserializers.BaseDeserializer): Optional.
+                Default parses responses from x-recordio-protobuf format.
         """
         super(NTMPredictor, self).__init__(
             endpoint_name,
             sagemaker_session,
-            serializer=RecordSerializer(),
-            deserializer=RecordDeserializer(),
+            serializer=serializer,
+            deserializer=deserializer,
         )
 
 

src/sagemaker/amazon/pca.py

@@ -193,12 +193,19 @@ class PCAPredictor(Predictor):
     to fit the model this Predictor performs inference on.
 
     :meth:`predict()` returns a list of
-    :class:`~sagemaker.amazon.record_pb2.Record` objects, one for each row in
+    :class:`~sagemaker.amazon.record_pb2.Record` objects (assuming the default
+    recordio-protobuf ``deserializer`` is used), one for each row in
     the input ``ndarray``. The lower dimension vector result is stored in the
     ``projection`` key of the ``Record.label`` field.
     """
 
-    def __init__(self, endpoint_name, sagemaker_session=None):
+    def __init__(
+        self,
+        endpoint_name,
+        sagemaker_session=None,
+        serializer=RecordSerializer(),
+        deserializer=RecordDeserializer(),
+    ):
         """
         Args:
             endpoint_name (str): Name of the Amazon SageMaker endpoint to which
@@ -207,12 +214,16 @@ def __init__(self, endpoint_name, sagemaker_session=None):
                 object, used for SageMaker interactions (default: None). If not
                 specified, one is created using the default AWS configuration
                 chain.
+            serializer (sagemaker.serializers.BaseSerializer): Optional. Default
+                serializes input data to x-recordio-protobuf format.
+            deserializer (sagemaker.deserializers.BaseDeserializer): Optional.
+                Default parses responses from x-recordio-protobuf format.
         """
         super(PCAPredictor, self).__init__(
             endpoint_name,
             sagemaker_session,
-            serializer=RecordSerializer(),
-            deserializer=RecordDeserializer(),
+            serializer=serializer,
+            deserializer=deserializer,
         )
 
 

src/sagemaker/amazon/randomcutforest.py

@@ -171,12 +171,19 @@ class RandomCutForestPredictor(Predictor):
     to fit the model this Predictor performs inference on.
 
     :meth:`predict()` returns a list of
-    :class:`~sagemaker.amazon.record_pb2.Record` objects, one for each row in
+    :class:`~sagemaker.amazon.record_pb2.Record` objects (assuming the default
+    recordio-protobuf ``deserializer`` is used), one for each row in
     the input. Each row's score is stored in the key ``score`` of the
     ``Record.label`` field.
     """
 
-    def __init__(self, endpoint_name, sagemaker_session=None):
+    def __init__(
+        self,
+        endpoint_name,
+        sagemaker_session=None,
+        serializer=RecordSerializer(),
+        deserializer=RecordDeserializer(),
+    ):
         """
         Args:
             endpoint_name (str): Name of the Amazon SageMaker endpoint to which
@@ -185,12 +192,16 @@ def __init__(self, endpoint_name, sagemaker_session=None):
                 object, used for SageMaker interactions (default: None). If not
                 specified, one is created using the default AWS configuration
                 chain.
+            serializer (sagemaker.serializers.BaseSerializer): Optional. Default
+                serializes input data to x-recordio-protobuf format.
+            deserializer (sagemaker.deserializers.BaseDeserializer): Optional.
+                Default parses responses from x-recordio-protobuf format.
         """
         super(RandomCutForestPredictor, self).__init__(
             endpoint_name,
             sagemaker_session,
-            serializer=RecordSerializer(),
-            deserializer=RecordDeserializer(),
+            serializer=serializer,
+            deserializer=deserializer,
         )
 
 

src/sagemaker/chainer/model.py

@@ -38,7 +38,13 @@ class ChainerPredictor(Predictor):
     multidimensional tensors for Chainer inference.
     """
 
-    def __init__(self, endpoint_name, sagemaker_session=None):
+    def __init__(
+        self,
+        endpoint_name,
+        sagemaker_session=None,
+        serializer=NumpySerializer(),
+        deserializer=NumpyDeserializer(),
+    ):
         """Initialize an ``ChainerPredictor``.
 
         Args:
@@ -48,9 +54,17 @@ def __init__(self, endpoint_name, sagemaker_session=None):
                 manages interactions with Amazon SageMaker APIs and any other
                 AWS services needed. If not specified, the estimator creates one
                 using the default AWS configuration chain.
+            serializer (sagemaker.serializers.BaseSerializer): Optional. Default
+                serializes input data to .npy format. Handles lists and numpy
+                arrays.
+            deserializer (sagemaker.deserializers.BaseDeserializer): Optional.
+                Default parses the response from .npy format to numpy array.
         """
         super(ChainerPredictor, self).__init__(
-            endpoint_name, sagemaker_session, NumpySerializer(), NumpyDeserializer()
+            endpoint_name,
+            sagemaker_session,
+            serializer=serializer,
+            deserializer=deserializer,
         )