Merge branch 'master' into master

Author: makungaj1

src/sagemaker/serve/builder/djl_builder.py

@@ -83,6 +83,7 @@ def __init__(self):
         self.mode = None
         self.model_server = None
         self.image_uri = None
+        self._is_custom_image_uri = False
         self.image_config = None
         self.vpc_config = None
         self._original_deploy = None

src/sagemaker/serve/builder/jumpstart_builder.py

@@ -57,6 +57,8 @@ def __init__(self):
         self.mode = None
         self.model_server = None
         self.image_uri = None
+        self._is_custom_image_uri = False
+        self.vpc_config = None
         self._original_deploy = None
         self.secret_key = None
         self.js_model_config = None
@@ -94,7 +96,7 @@ def _is_jumpstart_model_id(self) -> bool:
 
     def _create_pre_trained_js_model(self) -> Type[Model]:
         """Placeholder docstring"""
-        pysdk_model = JumpStartModel(self.model)
+        pysdk_model = JumpStartModel(self.model, vpc_config=self.vpc_config)
         pysdk_model.sagemaker_session = self.sagemaker_session
 
         self._original_deploy = pysdk_model.deploy

src/sagemaker/serve/builder/model_builder.py

@@ -607,6 +607,8 @@ def build(
 
         self.serve_settings = self._get_serve_setting()
 
+        self._is_custom_image_uri = self.image_uri is None
+
         if isinstance(self.model, str):
             if self._is_jumpstart_model_id():
                 return self._build_for_jumpstart()

src/sagemaker/serve/builder/tgi_builder.py

@@ -76,6 +76,7 @@ def __init__(self):
         self.mode = None
         self.model_server = None
         self.image_uri = None
+        self._is_custom_image_uri = False
         self.image_config = None
         self.vpc_config = None
         self._original_deploy = None

src/sagemaker/serve/builder/transformers_builder.py

@@ -56,6 +56,8 @@ def __init__(self):
         self.mode = None
         self.model_server = None
         self.image_uri = None
+        self._is_custom_image_uri = False
+        self.vpc_config = None
         self._original_deploy = None
         self.hf_model_config = None
         self._default_data_type = None
@@ -111,6 +113,7 @@ def _create_transformers_model(self) -> Type[Model]:
                 py_version=self.py_version,
                 transformers_version=base_hf_version,
                 pytorch_version=self.pytorch_version,
+                vpc_config=self.vpc_config,
             )
         elif "keras" in hf_model_md.get("tags") or "tensorflow" in hf_model_md.get("tags"):
             self.tensorflow_version = self._get_supported_version(
@@ -126,6 +129,7 @@ def _create_transformers_model(self) -> Type[Model]:
                 py_version=self.py_version,
                 transformers_version=base_hf_version,
                 tensorflow_version=self.tensorflow_version,
+                vpc_config=self.vpc_config,
             )
 
         if self.mode == Mode.LOCAL_CONTAINER:

src/sagemaker/serve/utils/telemetry_logger.py

@@ -20,7 +20,8 @@
 from sagemaker import Session, exceptions
 from sagemaker.serve.mode.function_pointers import Mode
 from sagemaker.serve.utils.exceptions import ModelBuilderException
-from sagemaker.serve.utils.types import ModelServer
+from sagemaker.serve.utils.types import ModelServer, ImageUriOption
+from sagemaker.serve.validations.check_image_uri import is_1p_image_uri
 from sagemaker.user_agent import SDK_VERSION
 
 logger = logging.getLogger(__name__)
@@ -62,11 +63,13 @@ def wrapper(self, *args, **kwargs):
             caught_ex = None
 
             image_uri_tail = self.image_uri.split("/")[1]
+            image_uri_option = _get_image_uri_option(self.image_uri, self._is_custom_image_uri)
             extra = (
                 f"{func_name}"
                 f"&x-modelServer={MODEL_SERVER_TO_CODE[str(self.model_server)]}"
                 f"&x-imageTag={image_uri_tail}"
                 f"&x-sdkVersion={SDK_VERSION}"
+                f"&x-defaultImageUsage={image_uri_option}"
             )
 
             if self.model_server == ModelServer.DJL_SERVING or self.model_server == ModelServer.TGI:
@@ -201,3 +204,22 @@ def _get_region_or_default(session):
         return session.boto_session.region_name
     except Exception:  # pylint: disable=W0703
         return "us-west-2"
+
+
+def _get_image_uri_option(image_uri: str, is_custom_image: bool) -> int:
+    """Detect whether default values are used for ModelBuilder
+
+    Args:
+        image_uri (str): Image uri used by ModelBuilder.
+        is_custom_image: (bool): Boolean indicating whether customer provides with custom image.
+    Returns:
+        bool: Integer code of image option types.
+    """
+
+    if not is_custom_image:
+        return ImageUriOption.DEFAULT_IMAGE.value
+
+    if is_1p_image_uri(image_uri):
+        return ImageUriOption.CUSTOM_1P_IMAGE.value
+
+    return ImageUriOption.CUSTOM_IMAGE.value

src/sagemaker/serve/utils/types.py

@@ -43,3 +43,15 @@ def __str__(self) -> str:
     INFERENTIA_1 = 3
     INFERENTIA_2 = 4
     GRAVITON = 5
+
+
+class ImageUriOption(Enum):
+    """Enum type for image uri options"""
+
+    def __str__(self) -> str:
+        """Convert enum to string"""
+        return str(self.name)
+
+    CUSTOM_IMAGE = 1
+    CUSTOM_1P_IMAGE = 2
+    DEFAULT_IMAGE = 3

src/sagemaker/workflow/entities.py

@@ -89,7 +89,7 @@ def __str__(self):
         """Override built-in String function for PipelineVariable"""
         raise TypeError(
             "Pipeline variables do not support __str__ operation. "
-            "Please use `.to_string()` to convert it to string type in execution time"
+            "Please use `.to_string()` to convert it to string type in execution time "
             "or use `.expr` to translate it to Json for display purpose in Python SDK."
         )
 

tests/unit/sagemaker/serve/builder/test_transformers_builder.py

@@ -16,6 +16,7 @@
 import unittest
 from sagemaker.serve.builder.model_builder import ModelBuilder
 from sagemaker.serve.mode.function_pointers import Mode
+from tests.unit.sagemaker.serve.constants import MOCK_VPC_CONFIG
 
 from sagemaker.serve.utils.predictors import TransformersLocalModePredictor
 
@@ -74,6 +75,7 @@ def test_build_deploy_for_transformers_local_container_and_remote_container(
             model=mock_model_id,
             schema_builder=mock_schema_builder,
             mode=Mode.LOCAL_CONTAINER,
+            vpc_config=MOCK_VPC_CONFIG,
         )
 
         builder._prepare_for_mode = MagicMock()
@@ -85,6 +87,7 @@ def test_build_deploy_for_transformers_local_container_and_remote_container(
         builder.modes[str(Mode.LOCAL_CONTAINER)] = MagicMock()
         predictor = model.deploy(model_data_download_timeout=1800)
 
+        assert model.vpc_config == MOCK_VPC_CONFIG
         assert builder.env_vars["MODEL_LOADING_TIMEOUT"] == "1800"
         assert isinstance(predictor, TransformersLocalModePredictor)
 

tests/unit/sagemaker/serve/utils/test_telemetry_logger.py

@@ -20,6 +20,7 @@
     _construct_url,
 )
 from sagemaker.serve.utils.exceptions import ModelBuilderException, LocalModelOutOfMemoryException
+from sagemaker.serve.utils.types import ImageUriOption
 from sagemaker.user_agent import SDK_VERSION
 
 MOCK_SESSION = Mock()
@@ -71,6 +72,7 @@ def test_capture_telemetry_decorator_djl_success(self, mock_send_telemetry):
         mock_model_builder = ModelBuilderMock()
         mock_model_builder.serve_settings.telemetry_opt_out = False
         mock_model_builder.image_uri = MOCK_DJL_CONTAINER
+        mock_model_builder._is_custom_image_uri = False
         mock_model_builder.model = MOCK_HUGGINGFACE_ID
         mock_model_builder.mode = Mode.LOCAL_CONTAINER
         mock_model_builder.model_server = ModelServer.DJL_SERVING
@@ -85,6 +87,37 @@ def test_capture_telemetry_decorator_djl_success(self, mock_send_telemetry):
             "&x-modelServer=4"
             "&x-imageTag=djl-inference:0.25.0-deepspeed0.11.0-cu118"
             f"&x-sdkVersion={SDK_VERSION}"
+            f"&x-defaultImageUsage={ImageUriOption.DEFAULT_IMAGE.value}"
+            f"&x-modelName={MOCK_HUGGINGFACE_ID}"
+            f"&x-endpointArn={MOCK_ENDPOINT_ARN}"
+            f"&x-latency={latency}"
+        )
+
+        mock_send_telemetry.assert_called_once_with(
+            "1", 2, MOCK_SESSION, None, None, expected_extra_str
+        )
+
+    @patch("sagemaker.serve.utils.telemetry_logger._send_telemetry")
+    def test_capture_telemetry_decorator_djl_success_with_custom_image(self, mock_send_telemetry):
+        mock_model_builder = ModelBuilderMock()
+        mock_model_builder.serve_settings.telemetry_opt_out = False
+        mock_model_builder.image_uri = MOCK_DJL_CONTAINER
+        mock_model_builder._is_custom_image_uri = True
+        mock_model_builder.model = MOCK_HUGGINGFACE_ID
+        mock_model_builder.mode = Mode.LOCAL_CONTAINER
+        mock_model_builder.model_server = ModelServer.DJL_SERVING
+        mock_model_builder.sagemaker_session.endpoint_arn = MOCK_ENDPOINT_ARN
+
+        mock_model_builder.mock_deploy()
+
+        args = mock_send_telemetry.call_args.args
+        latency = str(args[5]).split("latency=")[1]
+        expected_extra_str = (
+            f"{MOCK_FUNC_NAME}"
+            "&x-modelServer=4"
+            "&x-imageTag=djl-inference:0.25.0-deepspeed0.11.0-cu118"
+            f"&x-sdkVersion={SDK_VERSION}"
+            f"&x-defaultImageUsage={ImageUriOption.CUSTOM_1P_IMAGE.value}"
             f"&x-modelName={MOCK_HUGGINGFACE_ID}"
             f"&x-endpointArn={MOCK_ENDPOINT_ARN}"
             f"&x-latency={latency}"
@@ -99,6 +132,7 @@ def test_capture_telemetry_decorator_tgi_success(self, mock_send_telemetry):
         mock_model_builder = ModelBuilderMock()
         mock_model_builder.serve_settings.telemetry_opt_out = False
         mock_model_builder.image_uri = MOCK_TGI_CONTAINER
+        mock_model_builder._is_custom_image_uri = False
         mock_model_builder.model = MOCK_HUGGINGFACE_ID
         mock_model_builder.mode = Mode.LOCAL_CONTAINER
         mock_model_builder.model_server = ModelServer.TGI
@@ -113,6 +147,7 @@ def test_capture_telemetry_decorator_tgi_success(self, mock_send_telemetry):
             "&x-modelServer=6"
             "&x-imageTag=huggingface-pytorch-inference:2.0.0-transformers4.28.1-cpu-py310-ubuntu20.04"
             f"&x-sdkVersion={SDK_VERSION}"
+            f"&x-defaultImageUsage={ImageUriOption.DEFAULT_IMAGE.value}"
             f"&x-modelName={MOCK_HUGGINGFACE_ID}"
             f"&x-endpointArn={MOCK_ENDPOINT_ARN}"
             f"&x-latency={latency}"
@@ -127,6 +162,7 @@ def test_capture_telemetry_decorator_no_call_when_disabled(self, mock_send_telem
         mock_model_builder = ModelBuilderMock()
         mock_model_builder.serve_settings.telemetry_opt_out = True
         mock_model_builder.image_uri = MOCK_DJL_CONTAINER
+        mock_model_builder._is_custom_image_uri = False
         mock_model_builder.model = MOCK_HUGGINGFACE_ID
         mock_model_builder.model_server = ModelServer.DJL_SERVING
 
@@ -139,6 +175,7 @@ def test_capture_telemetry_decorator_handle_exception_success(self, mock_send_te
         mock_model_builder = ModelBuilderMock()
         mock_model_builder.serve_settings.telemetry_opt_out = False
         mock_model_builder.image_uri = MOCK_DJL_CONTAINER
+        mock_model_builder._is_custom_image_uri = False
         mock_model_builder.model = MOCK_HUGGINGFACE_ID
         mock_model_builder.mode = Mode.LOCAL_CONTAINER
         mock_model_builder.model_server = ModelServer.DJL_SERVING
@@ -158,6 +195,7 @@ def test_capture_telemetry_decorator_handle_exception_success(self, mock_send_te
             "&x-modelServer=4"
             "&x-imageTag=djl-inference:0.25.0-deepspeed0.11.0-cu118"
             f"&x-sdkVersion={SDK_VERSION}"
+            f"&x-defaultImageUsage={ImageUriOption.DEFAULT_IMAGE.value}"
             f"&x-modelName={MOCK_HUGGINGFACE_ID}"
             f"&x-endpointArn={MOCK_ENDPOINT_ARN}"
             f"&x-latency={latency}"