Merge branch 'master' into add-variant-name-kwarg

Author: mufaddal-rohawala

CHANGELOG.md

@@ -1,5 +1,78 @@
 # Changelog
 
+## v2.178.0 (2023-08-17)
+
+### Features
+
+ * Support to get latest monitoring execution processing logs
+
+### Bug Fixes and Other Changes
+
+ * Add context to predict_fn example
+ * gated models unsupported region
+ * jumpstart cache using sagemaker session s3 client
+ * add TFS 2.13 Graviton SM images
+ * pipeline variable kms key
+ * integration test for gated jumpstart training model
+ * tags for jumpstart model package models
+
+## v2.177.1 (2023-08-14)
+
+### Bug Fixes and Other Changes
+
+ * chore: excessive jumpstart bucket logging
+
+## v2.177.0 (2023-08-11)
+
+### Features
+
+ * Add TLV accounts for 1P Algorithms
+
+## v2.176.0 (2023-08-10)
+
+### Features
+
+ * Add TF 2.13 Training and Inference SM images
+
+### Bug Fixes and Other Changes
+
+ * revert-PR_3903
+ * skip tensorflow local mode notebook test
+ * change instance type for huggingface test to ml.g5.8xlarge
+
+## v2.175.0 (2023-08-05)
+
+### Features
+
+ * Add huggingface-llm 0.9.3 dlc images
+
+### Bug Fixes and Other Changes
+
+ * Upgrade default version for djl to v0.23.0
+ * Pass kms_key to _upload_analysis_config when provided
+
+## v2.174.0 (2023-08-02)
+
+### Features
+
+ * meta llama fine tuning
+ * support online store ttl for records
+ * Deploy uncompressed ML model from S3 to SageMaker Hosting endpoints
+ * AutoGluon 0.8.2 image_uris update
+
+### Bug Fixes and Other Changes
+
+ * [Feature] Propagate tags to lineage resources
+ * excessive jumpstart logging
+ * chore: jumpstart deprecation messages
+ * build(deps): bump pygments from 2.11.2 to 2.15.0 in /requirements/tox
+ * Remove deleted notebook tests from test confg
+ * chore: add jumpstart llama 2 tests
+
+### Documentation Changes
+
+ * add smp class for supporting flash attn
+
 ## v2.173.0 (2023-07-15)
 
 ### Features

VERSION

@@ -1 +1 @@
-2.173.1.dev0
+2.178.1.dev0

doc/frameworks/pytorch/using_pytorch.rst

@@ -772,7 +772,7 @@ The following example is for use cases with multiple GPUs and shows an overridde
     import torch
     import numpy as np
 
-    def predict_fn(input_data, model):
+    def predict_fn(input_data, model, context):
         device = torch.device("cuda:" + str(context.system_properties.get("gpu_id")) if torch.cuda.is_available() else "cpu")
         model.to(device)
         model.eval()

src/sagemaker/accept_types.py

@@ -15,6 +15,8 @@
 from typing import List, Optional
 
 from sagemaker.jumpstart import artifacts, utils as jumpstart_utils
+from sagemaker.jumpstart.constants import DEFAULT_JUMPSTART_SAGEMAKER_SESSION
+from sagemaker.session import Session
 
 
 def retrieve_options(
@@ -23,6 +25,7 @@ def retrieve_options(
     model_version: Optional[str] = None,
     tolerate_vulnerable_model: bool = False,
     tolerate_deprecated_model: bool = False,
+    sagemaker_session: Session = DEFAULT_JUMPSTART_SAGEMAKER_SESSION,
 ) -> List[str]:
     """Retrieves the supported accept types for the model matching the given arguments.
 
@@ -40,6 +43,10 @@ def retrieve_options(
         tolerate_deprecated_model (bool): True if deprecated models should be tolerated
             (exception not raised). False if these models should raise an exception.
             (Default: False).
+        sagemaker_session (sagemaker.session.Session): A SageMaker Session
+            object, used for SageMaker interactions. If not
+            specified, one is created using the default AWS configuration
+            chain. (Default: sagemaker.jumpstart.constants.DEFAULT_JUMPSTART_SAGEMAKER_SESSION).
     Returns:
         list: The supported accept types to use for the model.
 
@@ -57,6 +64,7 @@ def retrieve_options(
         region,
         tolerate_vulnerable_model,
         tolerate_deprecated_model,
+        sagemaker_session=sagemaker_session,
     )
 
 
@@ -66,6 +74,7 @@ def retrieve_default(
     model_version: Optional[str] = None,
     tolerate_vulnerable_model: bool = False,
     tolerate_deprecated_model: bool = False,
+    sagemaker_session: Session = DEFAULT_JUMPSTART_SAGEMAKER_SESSION,
 ) -> str:
     """Retrieves the default accept type for the model matching the given arguments.
 
@@ -83,6 +92,10 @@ def retrieve_default(
         tolerate_deprecated_model (bool): True if deprecated models should be tolerated
             (exception not raised). False if these models should raise an exception.
             (Default: False).
+        sagemaker_session (sagemaker.session.Session): A SageMaker Session
+            object, used for SageMaker interactions. If not
+            specified, one is created using the default AWS configuration
+            chain. (Default: sagemaker.jumpstart.constants.DEFAULT_JUMPSTART_SAGEMAKER_SESSION).
     Returns:
         str: The default accept type to use for the model.
 
@@ -100,4 +113,5 @@ def retrieve_default(
         region,
         tolerate_vulnerable_model,
         tolerate_deprecated_model,
+        sagemaker_session=sagemaker_session,
     )

src/sagemaker/content_types.py

@@ -15,6 +15,8 @@
 from typing import List, Optional
 
 from sagemaker.jumpstart import artifacts, utils as jumpstart_utils
+from sagemaker.jumpstart.constants import DEFAULT_JUMPSTART_SAGEMAKER_SESSION
+from sagemaker.session import Session
 
 
 def retrieve_options(
@@ -23,6 +25,7 @@ def retrieve_options(
     model_version: Optional[str] = None,
     tolerate_vulnerable_model: bool = False,
     tolerate_deprecated_model: bool = False,
+    sagemaker_session: Session = DEFAULT_JUMPSTART_SAGEMAKER_SESSION,
 ) -> List[str]:
     """Retrieves the supported content types for the model matching the given arguments.
 
@@ -40,6 +43,10 @@ def retrieve_options(
         tolerate_deprecated_model (bool): True if deprecated models should be tolerated
             (exception not raised). False if these models should raise an exception.
             (Default: False).
+        sagemaker_session (sagemaker.session.Session): A SageMaker Session
+            object, used for SageMaker interactions. If not
+            specified, one is created using the default AWS configuration
+            chain. (Default: sagemaker.jumpstart.constants.DEFAULT_JUMPSTART_SAGEMAKER_SESSION).
     Returns:
         list: The supported content types to use for the model.
 
@@ -57,6 +64,7 @@ def retrieve_options(
         region,
         tolerate_vulnerable_model,
         tolerate_deprecated_model,
+        sagemaker_session=sagemaker_session,
     )
 
 
@@ -66,6 +74,7 @@ def retrieve_default(
     model_version: Optional[str] = None,
     tolerate_vulnerable_model: bool = False,
     tolerate_deprecated_model: bool = False,
+    sagemaker_session: Session = DEFAULT_JUMPSTART_SAGEMAKER_SESSION,
 ) -> str:
     """Retrieves the default content type for the model matching the given arguments.
 
@@ -83,6 +92,10 @@ def retrieve_default(
         tolerate_deprecated_model (bool): True if deprecated models should be tolerated
             (exception not raised). False if these models should raise an exception.
             (Default: False).
+        sagemaker_session (sagemaker.session.Session): A SageMaker Session
+            object, used for SageMaker interactions. If not
+            specified, one is created using the default AWS configuration
+            chain. (Default: sagemaker.jumpstart.constants.DEFAULT_JUMPSTART_SAGEMAKER_SESSION).
     Returns:
         str: The default content type to use for the model.
 
@@ -100,6 +113,7 @@ def retrieve_default(
         region,
         tolerate_vulnerable_model,
         tolerate_deprecated_model,
+        sagemaker_session=sagemaker_session,
     )
 
 

src/sagemaker/deserializers.py

@@ -33,6 +33,8 @@
 )
 
 from sagemaker.jumpstart import artifacts, utils as jumpstart_utils
+from sagemaker.jumpstart.constants import DEFAULT_JUMPSTART_SAGEMAKER_SESSION
+from sagemaker.session import Session
 
 
 def retrieve_options(
@@ -41,6 +43,7 @@ def retrieve_options(
     model_version: Optional[str] = None,
     tolerate_vulnerable_model: bool = False,
     tolerate_deprecated_model: bool = False,
+    sagemaker_session: Session = DEFAULT_JUMPSTART_SAGEMAKER_SESSION,
 ) -> List[BaseDeserializer]:
     """Retrieves the supported deserializers for the model matching the given arguments.
 
@@ -58,6 +61,10 @@ def retrieve_options(
         tolerate_deprecated_model (bool): True if deprecated models should be tolerated
             (exception not raised). False if these models should raise an exception.
             (Default: False).
+        sagemaker_session (sagemaker.session.Session): A SageMaker Session
+            object, used for SageMaker interactions. If not
+            specified, one is created using the default AWS configuration
+            chain. (Default: sagemaker.jumpstart.constants.DEFAULT_JUMPSTART_SAGEMAKER_SESSION).
     Returns:
         List[BaseDeserializer]: The supported deserializers to use for the model.
 
@@ -76,6 +83,7 @@ def retrieve_options(
         region,
         tolerate_vulnerable_model,
         tolerate_deprecated_model,
+        sagemaker_session=sagemaker_session,
     )
 
 
@@ -85,6 +93,7 @@ def retrieve_default(
     model_version: Optional[str] = None,
     tolerate_vulnerable_model: bool = False,
     tolerate_deprecated_model: bool = False,
+    sagemaker_session: Session = DEFAULT_JUMPSTART_SAGEMAKER_SESSION,
 ) -> BaseDeserializer:
     """Retrieves the default deserializer for the model matching the given arguments.
 
@@ -102,6 +111,10 @@ def retrieve_default(
         tolerate_deprecated_model (bool): True if deprecated models should be tolerated
             (exception not raised). False if these models should raise an exception.
             (Default: False).
+        sagemaker_session (sagemaker.session.Session): A SageMaker Session
+            object, used for SageMaker interactions. If not
+            specified, one is created using the default AWS configuration
+            chain. (Default: sagemaker.jumpstart.constants.DEFAULT_JUMPSTART_SAGEMAKER_SESSION).
     Returns:
         BaseDeserializer: The default deserializer to use for the model.
 
@@ -120,4 +133,5 @@ def retrieve_default(
         region,
         tolerate_vulnerable_model,
         tolerate_deprecated_model,
+        sagemaker_session=sagemaker_session,
     )

src/sagemaker/djl_inference/model.py

@@ -781,7 +781,7 @@ def serving_image_uri(self, region_name):
             str: The appropriate image URI based on the given parameters.
         """
         if not self.djl_version:
-            self.djl_version = "0.22.1"
+            self.djl_version = "0.23.0"
 
         return image_uris.retrieve(
             self._framework(),

src/sagemaker/environment_variables.py

@@ -19,6 +19,8 @@
 
 from sagemaker.jumpstart import utils as jumpstart_utils
 from sagemaker.jumpstart import artifacts
+from sagemaker.jumpstart.constants import DEFAULT_JUMPSTART_SAGEMAKER_SESSION
+from sagemaker.session import Session
 
 logger = logging.getLogger(__name__)
 
@@ -30,6 +32,7 @@ def retrieve_default(
     tolerate_vulnerable_model: bool = False,
     tolerate_deprecated_model: bool = False,
     include_aws_sdk_env_vars: bool = True,
+    sagemaker_session: Session = DEFAULT_JUMPSTART_SAGEMAKER_SESSION,
 ) -> Dict[str, str]:
     """Retrieves the default container environment variables for the model matching the arguments.
 
@@ -51,6 +54,10 @@ def retrieve_default(
             should be included. The `Model` class of the SageMaker Python SDK inserts environment
             variables that would be required when making the low-level AWS API call.
             (Default: True).
+        sagemaker_session (sagemaker.session.Session): A SageMaker Session
+            object, used for SageMaker interactions. If not
+            specified, one is created using the default AWS configuration
+            chain. (Default: sagemaker.jumpstart.constants.DEFAULT_JUMPSTART_SAGEMAKER_SESSION).
     Returns:
         dict: The variables to use for the model.
 
@@ -70,4 +77,5 @@ def retrieve_default(
         tolerate_vulnerable_model,
         tolerate_deprecated_model,
         include_aws_sdk_env_vars,
+        sagemaker_session=sagemaker_session,
     )

src/sagemaker/estimator.py

@@ -99,7 +99,6 @@
 )
 from sagemaker.workflow import is_pipeline_variable
 from sagemaker.workflow.entities import PipelineVariable
-from sagemaker.workflow.parameters import ParameterString
 from sagemaker.workflow.pipeline_context import PipelineSession, runnable_by_pipeline
 
 logger = logging.getLogger(__name__)
@@ -614,16 +613,21 @@ def __init__(
         self.output_kms_key = resolve_value_from_config(
             output_kms_key, TRAINING_JOB_KMS_KEY_ID_PATH, sagemaker_session=self.sagemaker_session
         )
+        use_volume_kms_config: bool = False
         if instance_type is None or isinstance(instance_type, str):
             instance_type_for_volume_kms = instance_type
-        elif isinstance(instance_type, ParameterString):
-            instance_type_for_volume_kms = instance_type.default_value
+        elif isinstance(instance_type, PipelineVariable):
+            use_volume_kms_config = True
+            instance_type_for_volume_kms = instance_type
         else:
             raise ValueError(f"Bad value for instance type: '{instance_type}'")
 
         # KMS can only be attached to supported instances
         use_volume_kms_config = (
-            (instance_type_for_volume_kms and instance_supports_kms(instance_type_for_volume_kms))
+            use_volume_kms_config
+            or (
+                instance_type_for_volume_kms and instance_supports_kms(instance_type_for_volume_kms)
+            )
             or instance_groups is not None
             and any(
                 [
@@ -1425,6 +1429,24 @@ def attach(cls, training_job_name, sagemaker_session=None, model_channel_name="m
             Instance of the calling ``Estimator`` Class with the attached
             training job.
         """
+        return cls._attach(
+            training_job_name=training_job_name,
+            sagemaker_session=sagemaker_session,
+            model_channel_name=model_channel_name,
+        )
+
+    @classmethod
+    def _attach(
+        cls,
+        training_job_name: str,
+        sagemaker_session: Optional[str] = None,
+        model_channel_name: str = "model",
+        additional_kwargs: Optional[Dict[str, Any]] = None,
+    ) -> "EstimatorBase":
+        """Creates an Estimator bound to an existing training job.
+
+        Additional kwargs are allowed for instantiating Estimator.
+        """
         sagemaker_session = sagemaker_session or Session()
 
         job_details = sagemaker_session.sagemaker_client.describe_training_job(
@@ -1436,6 +1458,9 @@ def attach(cls, training_job_name, sagemaker_session=None, model_channel_name="m
         )["Tags"]
         init_params.update(tags=tags)
 
+        if additional_kwargs:
+            init_params.update(additional_kwargs)
+
         estimator = cls(sagemaker_session=sagemaker_session, **init_params)
         estimator.latest_training_job = _TrainingJob(
             sagemaker_session=sagemaker_session, job_name=training_job_name

src/sagemaker/huggingface/llm_utils.py

@@ -51,6 +51,6 @@ def get_huggingface_llm_image_uri(
             image_scope="inference",
         )
     if backend == "lmi":
-        version = version or "0.22.1"
+        version = version or "0.23.0"
         return image_uris.retrieve(framework="djl-deepspeed", region=region, version=version)
     raise ValueError("Unsupported backend: %s" % backend)