breaking: preserve script path when S3 source_dir is provided (#941)

Author: laurenyu

src/sagemaker/chainer/estimator.py

@@ -214,7 +214,7 @@ def create_model(
         return ChainerModel(
             self.model_data,
             role or self.role,
-            entry_point or self.entry_point,
+            entry_point or self._model_entry_point(),
             source_dir=(source_dir or self._model_source_dir()),
             enable_cloudwatch_metrics=self.enable_cloudwatch_metrics,
             container_log_level=self.container_log_level,

src/sagemaker/estimator.py

@@ -1734,17 +1734,28 @@ def _stage_user_code_in_s3(self):
         )
 
     def _model_source_dir(self):
-        """Get the appropriate value to pass as source_dir to model constructor
-        on deploying
+        """Get the appropriate value to pass as ``source_dir`` to a model constructor.
 
         Returns:
-            str: Either a local or an S3 path pointing to the source_dir to be
-            used for code by the model to be deployed
+            str: Either a local or an S3 path pointing to the ``source_dir`` to be
+                used for code by the model to be deployed
         """
         return (
             self.source_dir if self.sagemaker_session.local_mode else self.uploaded_code.s3_prefix
         )
 
+    def _model_entry_point(self):
+        """Get the appropriate value to pass as ``entry_point`` to a model constructor.
+
+        Returns:
+            str: The path to the entry point script. This can be either an absolute path or
+                a path relative to ``self._model_source_dir()``.
+        """
+        if self.sagemaker_session.local_mode or (self._model_source_dir() is None):
+            return self.entry_point
+
+        return self.uploaded_code.script_name
+
     def hyperparameters(self):
         """Return the hyperparameters as a dictionary to use for training.
 

src/sagemaker/fw_utils.py

@@ -447,7 +447,7 @@ def tar_and_upload_dir(
             script name.
     """
     if directory and directory.lower().startswith("s3://"):
-        return UploadedCode(s3_prefix=directory, script_name=os.path.basename(script))
+        return UploadedCode(s3_prefix=directory, script_name=script)
 
     script_name = script if directory else os.path.basename(script)
     dependencies = dependencies or []

src/sagemaker/mxnet/estimator.py

@@ -218,10 +218,10 @@ def create_model(
 
         kwargs["name"] = self._get_or_create_name(kwargs.get("name"))
 
-        return MXNetModel(
+        model = MXNetModel(
             self.model_data,
             role or self.role,
-            entry_point or self.entry_point,
+            entry_point,
             framework_version=self.framework_version,
             py_version=self.py_version,
             source_dir=(source_dir or self._model_source_dir()),
@@ -235,6 +235,13 @@ def create_model(
             **kwargs
         )
 
+        if entry_point is None:
+            model.entry_point = (
+                self.entry_point if model._is_mms_version() else self._model_entry_point()
+            )
+
+        return model
+
     @classmethod
     def _prepare_init_params_from_job_description(cls, job_details, model_channel_name=None):
         """Convert the job description to init params that can be handled by the

src/sagemaker/pytorch/estimator.py

@@ -175,7 +175,7 @@ def create_model(
         return PyTorchModel(
             self.model_data,
             role or self.role,
-            entry_point or self.entry_point,
+            entry_point or self._model_entry_point(),
             framework_version=self.framework_version,
             py_version=self.py_version,
             source_dir=(source_dir or self._model_source_dir()),

src/sagemaker/rl/estimator.py

@@ -232,7 +232,7 @@ def create_model(
         if not entry_point and (source_dir or dependencies):
             raise AttributeError("Please provide an `entry_point`.")
 
-        entry_point = entry_point or self.entry_point
+        entry_point = entry_point or self._model_entry_point()
         source_dir = source_dir or self._model_source_dir()
         dependencies = dependencies or self.dependencies
 

src/sagemaker/sklearn/estimator.py

@@ -196,7 +196,7 @@ def create_model(
         return SKLearnModel(
             self.model_data,
             role,
-            entry_point or self.entry_point,
+            entry_point or self._model_entry_point(),
             source_dir=(source_dir or self._model_source_dir()),
             enable_cloudwatch_metrics=self.enable_cloudwatch_metrics,
             container_log_level=self.container_log_level,

src/sagemaker/xgboost/estimator.py

@@ -172,7 +172,7 @@ def create_model(
         return XGBoostModel(
             self.model_data,
             role,
-            entry_point or self.entry_point,
+            entry_point or self._model_entry_point(),
             framework_version=self.framework_version,
             source_dir=(source_dir or self._model_source_dir()),
             enable_cloudwatch_metrics=self.enable_cloudwatch_metrics,

tests/data/mxnet_mnist/sourcedir.tar.gz

@@ -32,11 +32,16 @@ def mxnet_training_job(
     sagemaker_session, mxnet_full_version, mxnet_full_py_version, cpu_instance_type
 ):
     with timeout(minutes=TRAINING_DEFAULT_TIMEOUT_MINUTES):
-        script_path = os.path.join(DATA_DIR, "mxnet_mnist", "mnist.py")
+        s3_prefix = "integ-test-data/mxnet_mnist"
         data_path = os.path.join(DATA_DIR, "mxnet_mnist")
 
+        s3_source = sagemaker_session.upload_data(
+            path=os.path.join(data_path, "sourcedir.tar.gz"), key_prefix="{}/src".format(s3_prefix)
+        )
+
         mx = MXNet(
-            entry_point=script_path,
+            entry_point="mxnet_mnist/mnist.py",
+            source_dir=s3_source,
             role="SageMakerRole",
             framework_version=mxnet_full_version,
             py_version=mxnet_full_py_version,
@@ -46,10 +51,10 @@ def mxnet_training_job(
         )
 
         train_input = mx.sagemaker_session.upload_data(
-            path=os.path.join(data_path, "train"), key_prefix="integ-test-data/mxnet_mnist/train"
+            path=os.path.join(data_path, "train"), key_prefix="{}/train".format(s3_prefix)
         )
         test_input = mx.sagemaker_session.upload_data(
-            path=os.path.join(data_path, "test"), key_prefix="integ-test-data/mxnet_mnist/test"
+            path=os.path.join(data_path, "test"), key_prefix="{}/test".format(s3_prefix)
         )
 
         mx.fit({"train": train_input, "test": test_input})
@@ -62,7 +67,13 @@ def test_attach_deploy(mxnet_training_job, sagemaker_session, cpu_instance_type)
 
     with timeout_and_delete_endpoint_by_name(endpoint_name, sagemaker_session):
         estimator = MXNet.attach(mxnet_training_job, sagemaker_session=sagemaker_session)
-        predictor = estimator.deploy(1, cpu_instance_type, endpoint_name=endpoint_name)
+        predictor = estimator.deploy(
+            1,
+            cpu_instance_type,
+            entry_point="mnist.py",
+            source_dir=os.path.join(DATA_DIR, "mxnet_mnist"),
+            endpoint_name=endpoint_name,
+        )
         data = numpy.zeros(shape=(1, 1, 28, 28))
         result = predictor.predict(data)
         assert result is not None

tests/integ/test_mxnet.py

@@ -1269,7 +1269,7 @@ def test_git_support_codecommit_with_ssh_no_passphrase_needed(git_clone_repo, sa
 @patch("time.strftime", return_value=TIMESTAMP)
 def test_init_with_source_dir_s3(strftime, sagemaker_session):
     fw = DummyFramework(
-        entry_point=SCRIPT_PATH,
+        entry_point=SCRIPT_NAME,
         source_dir="s3://location",
         role=ROLE,
         sagemaker_session=sagemaker_session,

tests/unit/test_estimator.py

@@ -873,6 +873,18 @@ def test_tar_and_upload_dir_s3(sagemaker_session):
     assert result == fw_utils.UploadedCode("s3://m", "mnist.py")
 
 
+def test_tar_and_upload_dir_s3_with_script_dir(sagemaker_session):
+    bucket = "mybucket"
+    s3_key_prefix = "something/source"
+    script = "some/dir/mnist.py"
+    directory = "s3://m"
+    result = fw_utils.tar_and_upload_dir(
+        sagemaker_session, bucket, s3_key_prefix, script, directory
+    )
+
+    assert result == fw_utils.UploadedCode("s3://m", "some/dir/mnist.py")
+
+
 @patch("sagemaker.utils")
 def test_tar_and_upload_dir_s3_with_kms(utils, sagemaker_session):
     bucket = "mybucket"

tests/unit/test_fw_utils.py

@@ -27,7 +27,8 @@
 from sagemaker.mxnet import MXNetPredictor, MXNetModel
 
 DATA_DIR = os.path.join(os.path.dirname(__file__), "..", "data")
-SCRIPT_PATH = os.path.join(DATA_DIR, "dummy_script.py")
+SCRIPT_NAME = "dummy_script.py"
+SCRIPT_PATH = os.path.join(DATA_DIR, SCRIPT_NAME)
 SERVING_SCRIPT_FILE = "another_dummy_script.py"
 MODEL_DATA = "s3://mybucket/model"
 ENV = {"DUMMY_ENV_VAR": "dummy_value"}
@@ -189,7 +190,8 @@ def test_create_model(name_from_base, sagemaker_session, mxnet_version, mxnet_py
     base_job_name = "job"
 
     mx = MXNet(
-        entry_point=SCRIPT_PATH,
+        entry_point=SCRIPT_NAME,
+        source_dir=source_dir,
         framework_version=mxnet_version,
         py_version=mxnet_py_version,
         role=ROLE,
@@ -198,7 +200,6 @@ def test_create_model(name_from_base, sagemaker_session, mxnet_version, mxnet_py
         instance_type=INSTANCE_TYPE,
         container_log_level=container_log_level,
         base_job_name=base_job_name,
-        source_dir=source_dir,
     )
 
     mx.fit(inputs="s3://mybucket/train", job_name="new_name")
@@ -210,7 +211,7 @@ def test_create_model(name_from_base, sagemaker_session, mxnet_version, mxnet_py
     assert model.sagemaker_session == sagemaker_session
     assert model.framework_version == mxnet_version
     assert model.py_version == mxnet_py_version
-    assert model.entry_point == SCRIPT_PATH
+    assert model.entry_point == SCRIPT_NAME
     assert model.role == ROLE
     assert model.name == model_name
     assert model.container_log_level == container_log_level
@@ -226,7 +227,8 @@ def test_create_model_with_optional_params(sagemaker_session, mxnet_version, mxn
     source_dir = "s3://mybucket/source"
     enable_cloudwatch_metrics = "true"
     mx = MXNet(
-        entry_point=SCRIPT_PATH,
+        entry_point=SCRIPT_NAME,
+        source_dir=source_dir,
         framework_version=mxnet_version,
         py_version=mxnet_py_version,
         role=ROLE,
@@ -235,7 +237,6 @@ def test_create_model_with_optional_params(sagemaker_session, mxnet_version, mxn
         instance_type=INSTANCE_TYPE,
         container_log_level=container_log_level,
         base_job_name="job",
-        source_dir=source_dir,
         enable_cloudwatch_metrics=enable_cloudwatch_metrics,
     )
 
@@ -270,7 +271,8 @@ def test_create_model_with_custom_image(name_from_base, sagemaker_session):
     base_job_name = "job"
 
     mx = MXNet(
-        entry_point=SCRIPT_PATH,
+        entry_point=SCRIPT_NAME,
+        source_dir=source_dir,
         framework_version="2.0",
         py_version="py3",
         role=ROLE,
@@ -280,7 +282,6 @@ def test_create_model_with_custom_image(name_from_base, sagemaker_session):
         image_uri=custom_image,
         container_log_level=container_log_level,
         base_job_name=base_job_name,
-        source_dir=source_dir,
     )
 
     mx.fit(inputs="s3://mybucket/train", job_name="new_name")
@@ -291,7 +292,7 @@ def test_create_model_with_custom_image(name_from_base, sagemaker_session):
 
     assert model.sagemaker_session == sagemaker_session
     assert model.image_uri == custom_image
-    assert model.entry_point == SCRIPT_PATH
+    assert model.entry_point == SCRIPT_NAME
     assert model.role == ROLE
     assert model.name == model_name
     assert model.container_log_level == container_log_level
@@ -730,7 +731,6 @@ def test_model_py2_warning(warning, sagemaker_session):
 
 def test_create_model_with_custom_hosting_image(sagemaker_session):
     container_log_level = '"logging.INFO"'
-    source_dir = "s3://mybucket/source"
     custom_image = "mxnet:2.0"
     custom_hosting_image = "mxnet_hosting:2.0"
     mx = MXNet(
@@ -744,7 +744,6 @@ def test_create_model_with_custom_hosting_image(sagemaker_session):
         image_uri=custom_image,
         container_log_level=container_log_level,
         base_job_name="job",
-        source_dir=source_dir,
     )
 
     mx.fit(inputs="s3://mybucket/train", job_name="new_name")