Merge branch 'master' into update-pytorch

Author: sallyseok

CHANGELOG.md

@@ -1,5 +1,27 @@
 # Changelog
 
+## 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
+2.174.1.dev0

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/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)

src/sagemaker/jumpstart/artifacts/__init__.py

@@ -11,16 +11,21 @@
 # ANY KIND, either express or implied. See the License for the specific
 # language governing permissions and limitations under the License.
 """This module imports all JumpStart artifact functions from the respective sub-module."""
-from sagemaker.jumpstart.artifacts.prepack import _model_supports_prepacked_inference  # noqa: F401
 from sagemaker.jumpstart.artifacts.resource_names import (  # noqa: F401
     _retrieve_resource_name_base,
 )
 from sagemaker.jumpstart.artifacts.incremental_training import (  # noqa: F401
     _model_supports_incremental_training,
 )
 from sagemaker.jumpstart.artifacts.image_uris import _retrieve_image_uri  # noqa: F401
-from sagemaker.jumpstart.artifacts.script_uris import _retrieve_script_uri  # noqa: F401
-from sagemaker.jumpstart.artifacts.model_uris import _retrieve_model_uri  # noqa: F401
+from sagemaker.jumpstart.artifacts.script_uris import (  # noqa: F401
+    _retrieve_script_uri,
+    _model_supports_inference_script_uri,
+)
+from sagemaker.jumpstart.artifacts.model_uris import (  # noqa: F401
+    _retrieve_model_uri,
+    _model_supports_training_model_uri,
+)
 from sagemaker.jumpstart.artifacts.hyperparameters import (  # noqa: F401
     _retrieve_default_hyperparameters,
 )
@@ -52,4 +57,7 @@
     _retrieve_supported_accept_types,
     _retrieve_supported_content_types,
 )
-from sagemaker.jumpstart.artifacts.model_packages import _retrieve_model_package_arn  # noqa: F401
+from sagemaker.jumpstart.artifacts.model_packages import (  # noqa: F401
+    _retrieve_model_package_arn,
+    _retrieve_model_package_model_artifact_s3_uri,
+)

src/sagemaker/jumpstart/artifacts/model_packages.py

@@ -75,3 +75,61 @@ def _retrieve_model_package_arn(
         return regional_arn
 
     raise NotImplementedError(f"Model Package ARN not supported for scope: '{scope}'")
+
+
+def _retrieve_model_package_model_artifact_s3_uri(
+    model_id: str,
+    model_version: str,
+    region: Optional[str],
+    scope: Optional[str] = None,
+    tolerate_vulnerable_model: bool = False,
+    tolerate_deprecated_model: bool = False,
+) -> Optional[str]:
+    """Retrieves s3 artifact uri associated with model package.
+
+    Args:
+        model_id (str): JumpStart model ID of the JumpStart model for which to
+            retrieve the model package artifact.
+        model_version (str): Version of the JumpStart model for which to retrieve the
+            model package artifact.
+        region (Optional[str]): Region for which to retrieve the model package artifact.
+            (Default: None).
+        scope (Optional[str]): Scope for which to retrieve the model package artifact.
+            (Default: None).
+        tolerate_vulnerable_model (bool): True if vulnerable versions of model
+            specifications should be tolerated (exception not raised). If False, raises an
+            exception if the script used by this version of the model has dependencies with known
+            security vulnerabilities. (Default: False).
+        tolerate_deprecated_model (bool): True if deprecated versions of model
+            specifications should be tolerated (exception not raised). If False, raises
+            an exception if the version of the model is deprecated. (Default: False).
+
+    Returns:
+        str: the model package artifact uri to use for the model or None.
+
+    Raises:
+        NotImplementedError: If an unsupported script is used.
+    """
+
+    if scope == JumpStartScriptScope.TRAINING:
+
+        if region is None:
+            region = JUMPSTART_DEFAULT_REGION_NAME
+
+        model_specs = verify_model_region_and_return_specs(
+            model_id=model_id,
+            version=model_version,
+            scope=scope,
+            region=region,
+            tolerate_vulnerable_model=tolerate_vulnerable_model,
+            tolerate_deprecated_model=tolerate_deprecated_model,
+        )
+
+        if model_specs.training_model_package_artifact_uris is None:
+            return None
+
+        model_s3_uri = model_specs.training_model_package_artifact_uris.get(region)
+
+        return model_s3_uri
+
+    raise NotImplementedError(f"Model Package Artifact URI not supported for scope: '{scope}'")

src/sagemaker/jumpstart/artifacts/model_uris.py

@@ -92,3 +92,45 @@ def _retrieve_model_uri(
     model_s3_uri = f"s3://{bucket}/{model_artifact_key}"
 
     return model_s3_uri
+
+
+def _model_supports_training_model_uri(
+    model_id: str,
+    model_version: str,
+    region: Optional[str],
+    tolerate_vulnerable_model: bool = False,
+    tolerate_deprecated_model: bool = False,
+) -> bool:
+    """Returns True if the model supports training with model uri field.
+
+    Args:
+        model_id (str): JumpStart model ID of the JumpStart model for which to
+            retrieve the support status for model uri with training.
+        model_version (str): Version of the JumpStart model for which to retrieve the
+            support status for model uri with training.
+        region (Optional[str]): Region for which to retrieve the
+            support status for model uri with training.
+        tolerate_vulnerable_model (bool): True if vulnerable versions of model
+            specifications should be tolerated (exception not raised). If False, raises an
+            exception if the script used by this version of the model has dependencies with known
+            security vulnerabilities. (Default: False).
+        tolerate_deprecated_model (bool): True if deprecated versions of model
+            specifications should be tolerated (exception not raised). If False, raises
+            an exception if the version of the model is deprecated. (Default: False).
+    Returns:
+        bool: the support status for model uri with training.
+    """
+
+    if region is None:
+        region = JUMPSTART_DEFAULT_REGION_NAME
+
+    model_specs = verify_model_region_and_return_specs(
+        model_id=model_id,
+        version=model_version,
+        scope=JumpStartScriptScope.TRAINING,
+        region=region,
+        tolerate_vulnerable_model=tolerate_vulnerable_model,
+        tolerate_deprecated_model=tolerate_deprecated_model,
+    )
+
+    return model_specs.use_training_model_artifact()

src/sagemaker/jumpstart/artifacts/prepack.py

@@ -90,3 +90,45 @@ def _retrieve_script_uri(
     script_s3_uri = f"s3://{bucket}/{model_script_key}"
 
     return script_s3_uri
+
+
+def _model_supports_inference_script_uri(
+    model_id: str,
+    model_version: str,
+    region: Optional[str],
+    tolerate_vulnerable_model: bool = False,
+    tolerate_deprecated_model: bool = False,
+) -> bool:
+    """Returns True if the model supports inference with script uri field.
+
+    Args:
+        model_id (str): JumpStart model ID of the JumpStart model for which to
+            retrieve the support status for script uri with inference.
+        model_version (str): Version of the JumpStart model for which to retrieve the
+            support status for script uri with inference.
+        region (Optional[str]): Region for which to retrieve the
+            support status for script uri with inference.
+        tolerate_vulnerable_model (bool): True if vulnerable versions of model
+            specifications should be tolerated (exception not raised). If False, raises an
+            exception if the script used by this version of the model has dependencies with known
+            security vulnerabilities. (Default: False).
+        tolerate_deprecated_model (bool): True if deprecated versions of model
+            specifications should be tolerated (exception not raised). If False, raises
+            an exception if the version of the model is deprecated. (Default: False).
+    Returns:
+        bool: the support status for script uri with inference.
+    """
+
+    if region is None:
+        region = JUMPSTART_DEFAULT_REGION_NAME
+
+    model_specs = verify_model_region_and_return_specs(
+        model_id=model_id,
+        version=model_version,
+        scope=JumpStartScriptScope.INFERENCE,
+        region=region,
+        tolerate_vulnerable_model=tolerate_vulnerable_model,
+        tolerate_deprecated_model=tolerate_deprecated_model,
+    )
+
+    return model_specs.use_inference_script_uri()

src/sagemaker/jumpstart/artifacts/script_uris.py

@@ -147,6 +147,7 @@
 
 JUMPSTART_RESOURCE_BASE_NAME = "sagemaker-jumpstart"
 
+SAGEMAKER_GATED_MODEL_S3_URI_TRAINING_ENV_VAR_KEY = "SageMakerGatedModelS3Uri"
 
 CONTENT_TYPE_TO_SERIALIZER_TYPE_MAP: Dict[MIMEType, SerializerType] = {
     MIMEType.X_IMAGE: SerializerType.RAW_BYTES,