Merge branch 'master' into accelerate-source

Author: samruds

CHANGELOG.md

@@ -1,5 +1,31 @@
 # Changelog
 
+## v2.213.0 (2024-03-15)
+
+### Features
+
+ * Add support for Streaming Inference
+ * tgi optimum 0.0.19, 0.0.20 releases
+ * support JumpStart proprietary models
+ * Add ModelDataSource and SourceUri support for model package and while registering
+ * Accept user-defined env variables for the entry-point
+ * Add overriding logic in ModelBuilder when task is provided
+
+### Bug Fixes and Other Changes
+
+ * Improvement of the tuner documentation
+ * Skip of tests which are long running and causing the ResourceLimitInUse exception
+ * Add AutoML -> AutoMLV2 mapper
+ * add ci-health checks
+ * split coverage out from testenv in tox.ini
+ * add PT 2.2 support for smdistributed, pytorchddp, and torch_distributed distributions
+ * sagemaker session region not being used
+ * chore: emit warning when no instance specific gated training env var is available, and raise exception when accept_eula flag is not supplied
+ * enable github actions for PRs
+ * Move sagemaker pysdk version check after bootstrap in remote job
+ * make unit tests compatible with pytest-xdist
+ * Update tblib constraint
+
 ## v2.212.0 (2024-03-06)
 
 ### Features

README.rst

@@ -67,11 +67,10 @@ Table of Contents
 Installing the SageMaker Python SDK
 -----------------------------------
 
-The SageMaker Python SDK is built to PyPI and can be installed with pip as follows:
-
+The SageMaker Python SDK is built to PyPI and the latest version of the SageMaker Python SDK can be installed with pip as follows
 ::
 
-    pip install sagemaker
+    pip install sagemaker==<Latest version from pyPI from https://pypi.org/project/sagemaker/>
 
 You can install from source by cloning this repository and running a pip install command in the root directory of the repository:
 

VERSION

@@ -1 +1 @@
-2.212.1.dev0
+2.213.1.dev0

src/sagemaker/jumpstart/notebook_utils.py

@@ -262,6 +262,15 @@ def list_jumpstart_scripts(  # pylint: disable=redefined-builtin
     return sorted(list(scripts))
 
 
+def _is_valid_version(version: str) -> bool:
+    """Checks if the version is convertable to Version class."""
+    try:
+        Version(version)
+        return True
+    except Exception:  # pylint: disable=broad-except
+        return False
+
+
 def list_jumpstart_models(  # pylint: disable=redefined-builtin
     filter: Union[Operator, str] = Constant(BooleanValues.TRUE),
     region: Optional[str] = None,
@@ -304,7 +313,8 @@ def list_jumpstart_models(  # pylint: disable=redefined-builtin
     ):
         if model_id not in model_id_version_dict:
             model_id_version_dict[model_id] = list()
-        model_id_version_dict[model_id].append(Version(version))
+        model_version = Version(version) if _is_valid_version(version) else version
+        model_id_version_dict[model_id].append(model_version)
 
     if not list_versions:
         return sorted(list(model_id_version_dict.keys()))

src/sagemaker/tuner.py

@@ -11,6 +11,7 @@
 # ANY KIND, either express or implied. See the License for the specific
 # language governing permissions and limitations under the License.
 """Placeholder docstring"""
+
 from __future__ import absolute_import
 
 import importlib
@@ -641,8 +642,11 @@ def __init__(
                 extract the metric from the logs. This should be defined only
                 for hyperparameter tuning jobs that don't use an Amazon
                 algorithm.
-            strategy (str or PipelineVariable): Strategy to be used for hyperparameter estimations
-                (default: 'Bayesian').
+            strategy (str or PipelineVariable): Strategy to be used for hyperparameter estimations.
+                More information about different strategies:
+                https://docs.aws.amazon.com/sagemaker/latest/dg/automatic-model-tuning-how-it-works.html.
+                Available options are: 'Bayesian', 'Random', 'Hyperband',
+                'Grid' (default: 'Bayesian')
             objective_type (str or PipelineVariable): The type of the objective metric for
                 evaluating training jobs. This value can be either 'Minimize' or
                 'Maximize' (default: 'Maximize').
@@ -759,7 +763,8 @@ def __init__(
         self.autotune = autotune
 
     def override_resource_config(
-        self, instance_configs: Union[List[InstanceConfig], Dict[str, List[InstanceConfig]]]
+        self,
+        instance_configs: Union[List[InstanceConfig], Dict[str, List[InstanceConfig]]],
     ):
         """Override the instance configuration of the estimators used by the tuner.
 
@@ -966,7 +971,7 @@ def fit(
         include_cls_metadata: Union[bool, Dict[str, bool]] = False,
         estimator_kwargs: Optional[Dict[str, dict]] = None,
         wait: bool = True,
-        **kwargs
+        **kwargs,
     ):
         """Start a hyperparameter tuning job.
 
@@ -1055,7 +1060,7 @@ def _fit_with_estimator_dict(self, inputs, job_name, include_cls_metadata, estim
             allowed_keys=estimator_names,
         )
 
-        for (estimator_name, estimator) in self.estimator_dict.items():
+        for estimator_name, estimator in self.estimator_dict.items():
             ins = inputs.get(estimator_name, None) if inputs is not None else None
             args = estimator_kwargs.get(estimator_name, {}) if estimator_kwargs is not None else {}
             self._prepare_estimator_for_tuning(estimator, ins, job_name, **args)
@@ -1282,7 +1287,7 @@ def _attach_with_training_details_list(cls, sagemaker_session, estimator_cls, jo
             objective_metric_name_dict=objective_metric_name_dict,
             hyperparameter_ranges_dict=hyperparameter_ranges_dict,
             metric_definitions_dict=metric_definitions_dict,
-            **init_params
+            **init_params,
         )
 
     def deploy(
@@ -1297,7 +1302,7 @@ def deploy(
         model_name=None,
         kms_key=None,
         data_capture_config=None,
-        **kwargs
+        **kwargs,
     ):
         """Deploy the best trained or user specified model to an Amazon SageMaker endpoint.
 
@@ -1363,7 +1368,7 @@ def deploy(
             model_name=model_name,
             kms_key=kms_key,
             data_capture_config=data_capture_config,
-            **kwargs
+            **kwargs,
         )
 
     def stop_tuning_job(self):

tests/integ/sagemaker/serve/test_serve_pt_happy.py

@@ -181,6 +181,7 @@ def model_builder(request):
 #                 ), f"{caught_ex} was thrown when running pytorch squeezenet local container test"
 
 
+@pytest.mark.skip(reason="Failing test. Fix is pending.")
 @pytest.mark.skipif(
     PYTHON_VERSION_IS_NOT_310,  # or NOT_RUNNING_ON_INF_EXP_DEV_PIPELINE,
     reason="The goal of these test are to test the serving components of our feature",

tests/unit/sagemaker/jumpstart/constants.py

@@ -7577,6 +7577,20 @@
         "spec_key": "proprietary-models/ai21-paraphrase/proprietary_specs_1.0.005.json",
         "search_keywords": ["Text2Text", "Generation"],
     },
+    {
+        "model_id": "ai21-paraphrase",
+        "version": "v1.00-rc2-not-valid-version",
+        "min_version": "2.0.0",
+        "spec_key": "proprietary-models/ai21-paraphrase/proprietary_specs_1.0.005.json",
+        "search_keywords": ["Text2Text", "Generation"],
+    },
+    {
+        "model_id": "nc-soft-model-1",
+        "version": "v3.0-not-valid-version!",
+        "min_version": "2.0.0",
+        "spec_key": "proprietary-models/nc-soft-model-1/proprietary_specs_1.0.005.json",
+        "search_keywords": ["Text2Text", "Generation"],
+    },
 ]
 
 BASE_PROPRIETARY_SPEC = {

tests/unit/sagemaker/jumpstart/test_notebook_utils.py

@@ -25,6 +25,7 @@
     list_jumpstart_models,
     list_jumpstart_scripts,
     list_jumpstart_tasks,
+    _is_valid_version,
 )
 
 
@@ -185,6 +186,13 @@ def test_list_jumpstart_frameworks(
     patched_get_model_specs.assert_not_called()
 
 
+def test_is_valid_version():
+    valid_version_strs = ["1.0", "1.0.0", "2012.4", "1!1.0", "1.dev0", "1.2.3+abc.dev1"]
+    invalid_version_strs = ["1.1.053_m", "invalid version", "v1-1.0-v2", "@"]
+    assert all(_is_valid_version(v) for v in valid_version_strs)
+    assert not any(_is_valid_version(v) for v in invalid_version_strs)
+
+
 class ListJumpStartModels(TestCase):
     @patch("sagemaker.jumpstart.accessors.JumpStartModelsAccessor._get_manifest")
     @patch("sagemaker.jumpstart.accessors.JumpStartModelsAccessor.get_model_specs")
@@ -626,6 +634,7 @@ def test_list_jumpstart_proprietary_models(
             "ai21-paraphrase",
             "ai21-summarization",
             "lighton-mini-instruct40b",
+            "nc-soft-model-1",
         ]
 
         all_open_weight_model_ids = [