feature: Support for TFS preprocessing

src/sagemaker/local/image.py

@@ -235,10 +235,14 @@ def retrieve_artifacts(self, compose_data, output_data_config, job_name):
                     sagemaker.local.utils.recursive_copy(host_dir, output_artifacts)
 
         # Tar Artifacts -> model.tar.gz and output.tar.gz
-        model_files = [os.path.join(model_artifacts, name) for name in os.listdir(model_artifacts)]
-        output_files = [os.path.join(output_artifacts, name) for name in os.listdir(output_artifacts)]
-        sagemaker.utils.create_tar_file(model_files, os.path.join(compressed_artifacts, 'model.tar.gz'))
-        sagemaker.utils.create_tar_file(output_files, os.path.join(compressed_artifacts, 'output.tar.gz'))
+        model_files = [os.path.join(model_artifacts, name) for name in
+                       os.listdir(model_artifacts)]
+        output_files = [os.path.join(output_artifacts, name) for name in
+                        os.listdir(output_artifacts)]
+        sagemaker.utils.create_tar_file(model_files,
+                                        os.path.join(compressed_artifacts, 'model.tar.gz'))
+        sagemaker.utils.create_tar_file(output_files,
+                                        os.path.join(compressed_artifacts, 'output.tar.gz'))
 
         if output_data_config['S3OutputPath'] == '':
             output_data = 'file://%s' % compressed_artifacts

src/sagemaker/model.py

@@ -37,7 +37,7 @@ class Model(object):
     """A SageMaker ``Model`` that can be deployed to an ``Endpoint``."""
 
     def __init__(self, model_data, image, role=None, predictor_cls=None, env=None, name=None, vpc_config=None,
-                 sagemaker_session=None):
+                 sagemaker_session=None, enable_network_isolation=False):
         """Initialize an SageMaker ``Model``.
 
         Args:
@@ -58,6 +58,9 @@ def __init__(self, model_data, image, role=None, predictor_cls=None, env=None, n
                 * 'SecurityGroupIds' (list[str]): List of security group ids.
             sagemaker_session (sagemaker.session.Session): A SageMaker Session object, used for SageMaker
                interactions (default: None). If not specified, one is created using the default AWS configuration chain.
+            enable_network_isolation (Boolean): Default False. if True, enables network isolation in the endpoint,
+                isolating the model container. No inbound or outbound network calls can be made to or from the
+                model container.
         """
         self.model_data = model_data
         self.image = image
@@ -69,6 +72,7 @@ def __init__(self, model_data, image, role=None, predictor_cls=None, env=None, n
         self.sagemaker_session = sagemaker_session
         self._model_name = None
         self._is_compiled_model = False
+        self._enable_network_isolation = enable_network_isolation
 
     def prepare_container_def(self, instance_type, accelerator_type=None):  # pylint: disable=unused-argument
         """Return a dict created by ``sagemaker.container_def()`` for deploying this model to a specified instance type.
@@ -92,7 +96,7 @@ def enable_network_isolation(self):
         Returns:
             bool: If network isolation should be enabled or not.
         """
-        return False
+        return self._enable_network_isolation
 
     def _create_sagemaker_model(self, instance_type, accelerator_type=None, tags=None):
         """Create a SageMaker Model Entity

src/sagemaker/tensorflow/serving.py

@@ -13,6 +13,7 @@
 from __future__ import absolute_import
 
 import logging
+
 import sagemaker
 from sagemaker.content_types import CONTENT_TYPE_JSON
 from sagemaker.fw_utils import create_image_uri
@@ -88,7 +89,7 @@ def predict(self, data, initial_args=None):
         return super(Predictor, self).predict(data, args)
 
 
-class Model(sagemaker.Model):
+class Model(sagemaker.model.FrameworkModel):
     FRAMEWORK_NAME = 'tensorflow-serving'
     LOG_LEVEL_PARAM_NAME = 'SAGEMAKER_TFS_NGINX_LOGLEVEL'
     LOG_LEVEL_MAP = {
@@ -99,7 +100,7 @@ class Model(sagemaker.Model):
         logging.CRITICAL: 'crit',
     }
 
-    def __init__(self, model_data, role, image=None, framework_version=TF_VERSION,
+    def __init__(self, model_data, role, entry_point=None, image=None, framework_version=TF_VERSION,
                  container_log_level=None, predictor_cls=Predictor, **kwargs):
         """Initialize a Model.
 
@@ -118,14 +119,23 @@ def __init__(self, model_data, role, image=None, framework_version=TF_VERSION,
            **kwargs: Keyword arguments passed to the ``Model`` initializer.
        """
         super(Model, self).__init__(model_data=model_data, role=role, image=image,
-                                    predictor_cls=predictor_cls, **kwargs)
+                                    predictor_cls=predictor_cls, entry_point=entry_point, **kwargs)
         self._framework_version = framework_version
         self._container_log_level = container_log_level
 
     def prepare_container_def(self, instance_type, accelerator_type=None):
         image = self._get_image_uri(instance_type, accelerator_type)
         env = self._get_container_env()
-        return sagemaker.container_def(image, self.model_data, env)
+
+        if self.entry_point:
+            model_data = sagemaker.utils.repack_model(self.entry_point,
+                                                      self.source_dir,
+                                                      self.model_data,
+                                                      self.sagemaker_session)
+        else:
+            model_data = self.model_data
+
+        return sagemaker.container_def(image, model_data, env)
 
     def _get_container_env(self):
         if not self._container_log_level:

src/sagemaker/utils.py

@@ -12,20 +12,24 @@
 # language governing permissions and limitations under the License.
 from __future__ import absolute_import
 
+import contextlib
 import errno
 import os
 import random
 import re
+import shutil
 import sys
 import tarfile
 import tempfile
 import time
 
 from datetime import datetime
 from functools import wraps
+from six.moves.urllib import parse
 
 import six
 
+import sagemaker
 
 ECR_URI_PATTERN = r'^(\d+)(\.)dkr(\.)ecr(\.)(.+)(\.)(amazonaws.com|c2s.ic.gov)(/)(.*:.*)$'
 
@@ -258,13 +262,10 @@ def download_folder(bucket_name, prefix, target, sagemaker_session):
 
 def create_tar_file(source_files, target=None):
     """Create a tar file containing all the source_files
-
     Args:
         source_files (List[str]): List of file paths that will be contained in the tar file
-
     Returns:
          (str): path to created tar file
-
     """
     if target:
         filename = target
@@ -278,6 +279,100 @@ def create_tar_file(source_files, target=None):
     return filename
 
 
+@contextlib.contextmanager
+def _tmpdir(suffix='', prefix='tmp'):
+    """Create a temporary directory with a context manager. The file is deleted when the context exits.
+
+    The prefix, suffix, and dir arguments are the same as for mkstemp().
+
+    Args:
+        suffix (str):  If suffix is specified, the file name will end with that suffix, otherwise there will be no
+                        suffix.
+        prefix (str):  If prefix is specified, the file name will begin with that prefix; otherwise,
+                        a default prefix is used.
+        dir (str):  If dir is specified, the file will be created in that directory; otherwise, a default directory is
+                        used.
+    Returns:
+        str: path to the directory
+    """
+    tmp = tempfile.mkdtemp(suffix=suffix, prefix=prefix, dir=None)
+    yield tmp
+    shutil.rmtree(tmp)
+
+
+def repack_model(inference_script, source_directory, model_uri, sagemaker_session):
+    """Unpack model tarball and creates a new model tarball with the provided code script.
+
+    This function does the following:
+    - uncompresses model tarball from S3 or local system into a temp folder
+    - replaces the inference code from the model with the new code provided
+    - compresses the new model tarball and saves it in S3 or local file system
+
+    Args:
+        inference_script (str): path or basename of the inference script that will be packed into the model
+        source_directory (str): path including all the files that will be packed into the model
+        model_uri (str): S3 or file system location of the original model tar
+        sagemaker_session (:class:`sagemaker.session.Session`): a sagemaker session to interact with S3.
+
+    Returns:
+        str: path to the new packed model
+    """
+    new_model_name = 'model-%s.tar.gz' % sagemaker.utils.sagemaker_short_timestamp()
+
+    with _tmpdir() as tmp:
+        tmp_model_dir = os.path.join(tmp, 'model')
+        os.mkdir(tmp_model_dir)
+
+        model_from_s3 = model_uri.startswith('s3://')
+        if model_from_s3:
+            local_model_path = os.path.join(tmp, 'tar_file')
+            download_file_from_url(model_uri, local_model_path, sagemaker_session)
+
+            new_model_path = os.path.join(tmp, new_model_name)
+        else:
+            local_model_path = model_uri.replace('file://', '')
+            new_model_path = os.path.join(os.path.dirname(local_model_path), new_model_name)
+
+        with tarfile.open(name=local_model_path, mode='r:gz') as t:
+            t.extractall(path=tmp_model_dir)
+
+        code_dir = os.path.join(tmp_model_dir, 'code')
+        if os.path.exists(code_dir):
+            shutil.rmtree(code_dir, ignore_errors=True)
+
+        dirname = source_directory if source_directory else os.path.dirname(inference_script)
+
+        shutil.copytree(dirname, code_dir)
+
+        with tarfile.open(new_model_path, mode='w:gz') as t:
+            t.add(tmp_model_dir, arcname=os.path.sep)
+
+        if model_from_s3:
+            url = parse.urlparse(model_uri)
+            bucket, key = url.netloc, url.path.lstrip('/')
+            new_key = key.replace(os.path.basename(key), new_model_name)
+
+            sagemaker_session.boto_session.resource('s3').Object(bucket, new_key).upload_file(new_model_path)
+            return 's3://%s/%s' % (bucket, new_key)
+        else:
+            return 'file://%s' % new_model_path
+
+
+def _list_files(script, directory):
+    if directory is None:
+        return [script]
+
+    basedir = directory if directory else os.path.dirname(script)
+    return [os.path.join(basedir, name) for name in os.listdir(basedir)]
+
+
+def download_file_from_url(url, dst, sagemaker_session):
+    url = parse.urlparse(url)
+    bucket, key = url.netloc, url.path.lstrip('/')
+
+    download_file(bucket, key, dst, sagemaker_session)
+
+
 def download_file(bucket_name, path, target, sagemaker_session):
     """Download a Single File from S3 into a local path
 

tests/data/tfs/tfs-test-model-with-inference/00000123/assets/foo.txt

@@ -0,0 +1 @@
+asset-file-contents

tests/data/tfs/tfs-test-model-with-inference/code/inference.py

@@ -0,0 +1,26 @@
+# Copyright 2018 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.
+import json
+
+
+def input_handler(data, context):
+    data = json.loads(data.read().decode('utf-8'))
+    new_values = [x + 1 for x in data['instances']]
+    dumps = json.dumps({'instances': new_values})
+    return dumps
+
+
+def output_handler(data, context):
+    response_content_type = context.accept_header
+    prediction = data.content
+    return prediction, response_content_type