Resolve merge conflicts

Author: Balaji Veeramani

MANIFEST.in

@@ -1,4 +1,6 @@
-recursive-include src/sagemaker *
+recursive-include src/sagemaker *.py
+
+include src/sagemaker/image_uri_config/*.json
 
 include VERSION
 include LICENSE.txt

doc/frameworks/tensorflow/upgrade_from_legacy.rst

@@ -245,10 +245,10 @@ For example, if you want to use JSON serialization and deserialization:
 
 .. code:: python
 
-    from sagemaker.predictor import json_deserializer, json_serializer
+    from sagemaker.predictor import json_deserializer
+    from sagemaker.serializers import JSONSerializer
 
-    predictor.content_type = "application/json"
-    predictor.serializer = json_serializer
+    predictor.serializer = JSONSerializer()
     predictor.accept = "application/json"
     predictor.deserializer = json_deserializer
 

setup.py

@@ -83,6 +83,7 @@ def read_version():
     packages=find_packages("src"),
     package_dir={"": "src"},
     py_modules=[os.path.splitext(os.path.basename(path))[0] for path in glob("src/*.py")],
+    include_package_data=True,
     long_description=read("README.rst"),
     author="Amazon Web Services",
     url="https://github.com/aws/sagemaker-python-sdk/",

src/sagemaker/cli/compatibility/v2/modifiers/framework_version.py

@@ -15,6 +15,8 @@
 
 import ast
 
+from packaging.version import InvalidVersion, Version
+
 from sagemaker.cli.compatibility.v2.modifiers import matching, parsing
 from sagemaker.cli.compatibility.v2.modifiers.modifier import Modifier
 
@@ -135,10 +137,15 @@ def _tf_py_version_default(framework_version):
     """Gets the py_version default based on framework_version for TensorFlow."""
     if not framework_version:
         return "py2"
-    version = [int(s) for s in framework_version.split(".")]
-    if version < [1, 12]:
+
+    try:
+        version = Version(framework_version)
+    except InvalidVersion:
+        return "py2"
+
+    if version < Version("1.12"):
         return "py2"
-    if version < [2, 2]:
+    if version < Version("2.2"):
         return "py3"
     return "py37"
 
@@ -186,7 +193,6 @@ def _version_args_needed(node):
     framework, is_model = _framework_from_node(node)
     expecting_py_version = _py_version_defaults(framework, framework_version, is_model)
     if expecting_py_version:
-        py_version = parsing.arg_value(node, PY_ARG)
-        return py_version is None
+        return not matching.has_arg(node, PY_ARG)
 
     return False

src/sagemaker/cli/compatibility/v2/modifiers/tf_legacy_mode.py

@@ -79,9 +79,11 @@ def _is_legacy_mode(self, node):
 
         for kw in node.keywords:
             if kw.arg == "script_mode":
-                script_mode = bool(kw.value.value)
+                script_mode = (
+                    bool(kw.value.value) if isinstance(kw.value, ast.NameConstant) else True
+                )
             if kw.arg == "py_version":
-                py_version = kw.value.s
+                py_version = kw.value.s if isinstance(kw.value, ast.Str) else "py3"
 
         return not (py_version.startswith("py3") or script_mode)
 
@@ -124,7 +126,8 @@ def modify_node(self, node):
 
         if add_image_uri:
             image_uri = self._image_uri_from_args(node.keywords)
-            node.keywords.append(ast.keyword(arg="image_uri", value=ast.Str(s=image_uri)))
+            if image_uri:
+                node.keywords.append(ast.keyword(arg="image_uri", value=ast.Str(s=image_uri)))
 
         node.keywords.append(ast.keyword(arg="model_dir", value=ast.NameConstant(value=False)))
 
@@ -155,19 +158,22 @@ def _to_ast_keyword(self, hps):
         return None
 
     def _image_uri_from_args(self, keywords):
-        """Returns a legacy TensorFlow image URI based on the estimator arguments."""
+        """Returns a legacy TensorFlow image URI based on the estimator arguments if possible."""
         tf_version = framework_version.FRAMEWORK_DEFAULTS["TensorFlow"]
         instance_type = "ml.m4.xlarge"  # CPU default (exact type doesn't matter)
 
         for kw in keywords:
             if kw.arg == "framework_version":
-                tf_version = kw.value.s
+                tf_version = kw.value.s if isinstance(kw.value, ast.Str) else None
             if kw.arg == "train_instance_type":
-                instance_type = kw.value.s
+                instance_type = kw.value.s if isinstance(kw.value, ast.Str) else None
 
-        return fw_utils.create_image_uri(
-            self.region, "tensorflow", instance_type, tf_version, "py2"
-        )
+        if tf_version and instance_type:
+            return fw_utils.create_image_uri(
+                self.region, "tensorflow", instance_type, tf_version, "py2"
+            )
+
+        return None
 
 
 class TensorBoardParameterRemover(Modifier):

src/sagemaker/deserializers.py

@@ -13,6 +13,8 @@
 """Implements methods for deserializing data returned from an inference endpoint."""
 from __future__ import absolute_import
 
+import csv
+
 import abc
 import codecs
 import io
@@ -96,6 +98,37 @@ def deserialize(self, data, content_type):
             data.close()
 
 
+class CSVDeserializer(BaseDeserializer):
+    """Deserialize a stream of bytes into a list of lists."""
+
+    ACCEPT = "text/csv"
+
+    def __init__(self, encoding="utf-8"):
+        """Initialize the string encoding.
+
+        Args:
+            encoding (str): The string encoding to use (default: "utf-8").
+        """
+        self.encoding = encoding
+
+    def deserialize(self, data, content_type):
+        """Deserialize data from an inference endpoint into a list of lists.
+
+        Args:
+            data (botocore.response.StreamingBody): Data to be deserialized.
+            content_type (str): The MIME type of the data.
+
+        Returns:
+            list: The data deserialized into a list of lists representing the
+                contents of a CSV file.
+        """
+        try:
+            decoded_string = data.read().decode(self.encoding)
+            return list(csv.reader(decoded_string.splitlines()))
+        finally:
+            data.close()
+
+
 class StreamDeserializer(BaseDeserializer):
     """Returns the data and content-type received from an inference endpoint.
 

src/sagemaker/image_uri_config/chainer.json

@@ -0,0 +1,94 @@
+{
+    "processors": ["cpu", "gpu"],
+    "version_aliases": {
+        "4.0": "4.0.0",
+        "4.1": "4.1.0",
+        "5.0": "5.0.0"
+    },
+    "versions": {
+        "4.0.0": {
+            "registries": {
+                "ap-east-1": "057415533634",
+                "ap-northeast-1": "520713654638",
+                "ap-northeast-2": "520713654638",
+                "ap-south-1": "520713654638",
+                "ap-southeast-1": "520713654638",
+                "ap-southeast-2": "520713654638",
+                "ca-central-1": "520713654638",
+                "cn-north-1": "422961961927",
+                "cn-northwest-1": "423003514399",
+                "eu-central-1": "520713654638",
+                "eu-north-1": "520713654638",
+                "eu-west-1": "520713654638",
+                "eu-west-2": "520713654638",
+                "eu-west-3": "520713654638",
+                "me-south-1": "724002660598",
+                "sa-east-1": "520713654638",
+                "us-east-1": "520713654638",
+                "us-east-2": "520713654638",
+                "us-gov-west-1": "246785580436",
+                "us-iso-east-1": "744548109606",
+                "us-west-1": "520713654638",
+                "us-west-2": "520713654638"
+            },
+            "repository": "sagemaker-chainer",
+            "py_versions": ["py2", "py3"]
+        },
+        "4.1.0": {
+            "registries": {
+                "ap-east-1": "057415533634",
+                "ap-northeast-1": "520713654638",
+                "ap-northeast-2": "520713654638",
+                "ap-south-1": "520713654638",
+                "ap-southeast-1": "520713654638",
+                "ap-southeast-2": "520713654638",
+                "ca-central-1": "520713654638",
+                "cn-north-1": "422961961927",
+                "cn-northwest-1": "423003514399",
+                "eu-central-1": "520713654638",
+                "eu-north-1": "520713654638",
+                "eu-west-1": "520713654638",
+                "eu-west-2": "520713654638",
+                "eu-west-3": "520713654638",
+                "me-south-1": "724002660598",
+                "sa-east-1": "520713654638",
+                "us-east-1": "520713654638",
+                "us-east-2": "520713654638",
+                "us-gov-west-1": "246785580436",
+                "us-iso-east-1": "744548109606",
+                "us-west-1": "520713654638",
+                "us-west-2": "520713654638"
+            },
+            "repository": "sagemaker-chainer",
+            "py_versions": ["py2", "py3"]
+        },
+        "5.0.0": {
+            "registries": {
+                "ap-east-1": "057415533634",
+                "ap-northeast-1": "520713654638",
+                "ap-northeast-2": "520713654638",
+                "ap-south-1": "520713654638",
+                "ap-southeast-1": "520713654638",
+                "ap-southeast-2": "520713654638",
+                "ca-central-1": "520713654638",
+                "cn-north-1": "422961961927",
+                "cn-northwest-1": "423003514399",
+                "eu-central-1": "520713654638",
+                "eu-north-1": "520713654638",
+                "eu-west-1": "520713654638",
+                "eu-west-2": "520713654638",
+                "eu-west-3": "520713654638",
+                "me-south-1": "724002660598",
+                "sa-east-1": "520713654638",
+                "us-east-1": "520713654638",
+                "us-east-2": "520713654638",
+                "us-gov-west-1": "246785580436",
+                "us-iso-east-1": "744548109606",
+                "us-west-1": "520713654638",
+                "us-west-2": "520713654638"
+            },
+            "repository": "sagemaker-chainer",
+            "py_versions": ["py2", "py3"]
+        }
+    }
+}

src/sagemaker/image_uris.py

@@ -0,0 +1,127 @@
+# Copyright 2020 Amazon.com, Inc. or its affiliates. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License"). You
+# may not use this file except in compliance with the License. A copy of
+# the License is located at
+#
+#     http://aws.amazon.com/apache2.0/
+#
+# or in the "license" file accompanying this file. This file is
+# distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF
+# ANY KIND, either express or implied. See the License for the specific
+# language governing permissions and limitations under the License.
+"""Functions for generating ECR image URIs for pre-built SageMaker Docker images."""
+from __future__ import absolute_import
+
+import json
+import os
+
+from sagemaker import utils
+
+ECR_URI_TEMPLATE = "{registry}.dkr.{hostname}/{repository}:{tag}"
+
+
+def retrieve(framework, region, version=None, py_version=None, instance_type=None):
+    """Retrieves the ECR URI for the Docker image matching the given arguments.
+
+    Args:
+        framework (str): The name of the framework.
+        region (str): The AWS region.
+        version (str): The framework version. This is required if there is
+            more than one supported version for the given framework.
+        py_version (str): The Python version. This is required if there is
+            more than one supported Python version for the given framework version.
+        instance_type (str): The SageMaker instance type. For supported types, see
+            https://aws.amazon.com/sagemaker/pricing/instance-types. This is required if
+            there are different images for different processor types.
+
+    Returns:
+        str: the ECR URI for the corresponding SageMaker Docker image.
+
+    Raises:
+        ValueError: If the framework version, Python version, processor type, or region is
+            not supported given the other arguments.
+    """
+    config = config_for_framework(framework)
+    version_config = config["versions"][_version_for_config(version, config, framework)]
+
+    registry = _registry_from_region(region, version_config["registries"])
+    hostname = utils._botocore_resolver().construct_endpoint("ecr", region)["hostname"]
+
+    repo = version_config["repository"]
+
+    _validate_py_version(py_version, version_config["py_versions"], framework, version)
+    tag = "{}-{}-{}".format(version, _processor(instance_type, config["processors"]), py_version)
+
+    return ECR_URI_TEMPLATE.format(registry=registry, hostname=hostname, repository=repo, tag=tag)
+
+
+def config_for_framework(framework):
+    """Loads the JSON config for the given framework."""
+    fname = os.path.join(os.path.dirname(__file__), "image_uri_config", "{}.json".format(framework))
+    with open(fname) as f:
+        return json.load(f)
+
+
+def _version_for_config(version, config, framework):
+    """Returns the version string for retrieving a framework version's specific config."""
+    if "version_aliases" in config:
+        if version in config["version_aliases"].keys():
+            return config["version_aliases"][version]
+
+    available_versions = config["versions"].keys()
+    if version in available_versions:
+        return version
+
+    raise ValueError(
+        "Unsupported {} version: {}. "
+        "You may need to upgrade your SDK version (pip install -U sagemaker) for newer versions. "
+        "Supported version(s): {}.".format(framework, version, ", ".join(available_versions))
+    )
+
+
+def _registry_from_region(region, registry_dict):
+    """Returns the ECR registry (AWS account number) for the given region."""
+    available_regions = registry_dict.keys()
+    if region not in available_regions:
+        raise ValueError(
+            "Unsupported region: {}. You may need to upgrade "
+            "your SDK version (pip install -U sagemaker) for newer regions. "
+            "Supported region(s): {}.".format(region, ", ".join(available_regions))
+        )
+
+    return registry_dict[region]
+
+
+def _processor(instance_type, available_processors):
+    """Returns the processor type for the given instance type."""
+    if instance_type.startswith("local"):
+        processor = "cpu" if instance_type == "local" else "gpu"
+    elif not instance_type.startswith("ml."):
+        raise ValueError(
+            "Invalid SageMaker instance type: {}. See: "
+            "https://aws.amazon.com/sagemaker/pricing/instance-types".format(instance_type)
+        )
+    else:
+        family = instance_type.split(".")[1]
+        processor = "gpu" if family[0] in ("g", "p") else "cpu"
+
+    if processor in available_processors:
+        return processor
+
+    raise ValueError(
+        "Unsupported processor type: {} (for {}). "
+        "Supported type(s): {}.".format(processor, instance_type, ", ".join(available_processors))
+    )
+
+
+def _validate_py_version(py_version, available_versions, framework, fw_version):
+    """Checks if the Python version is one of the supported versions."""
+    if py_version not in available_versions:
+        raise ValueError(
+            "Unsupported Python version for {} {}: {}. You may need to upgrade "
+            "your SDK version (pip install -U sagemaker) for newer versions. "
+            "Supported Python version(s): {}.".format(
+                framework, fw_version, py_version, ", ".join(available_versions)
+            )
+        )