Merge branch 'master' into mwfongAWS-SM-sdk

Author: mwfongAWS

CHANGELOG.md

@@ -1,5 +1,16 @@
 # Changelog
 
+## v2.155.0 (2023-05-15)
+
+### Features
+
+ * Add support for SageMaker Serverless inference Provisioned Concurrency feature
+
+### Bug Fixes and Other Changes
+
+ * Revert "fix: make RemoteExecutor context manager non-blocking on pend…
+ * Add BOM to no No P2 Availability region list
+
 ## v2.154.0 (2023-05-11)
 
 ### Features

VERSION

@@ -1 +1 @@
-2.154.1.dev0
+2.155.1.dev0

src/sagemaker/djl_inference/model.py

@@ -854,11 +854,13 @@ def generate_serving_properties(self, serving_properties=None) -> Dict[str, str]
         if self.low_cpu_mem_usage:
             serving_properties["option.low_cpu_mem_usage"] = self.low_cpu_mem_usage
         # This is a workaround due to a bug in our built in handler for huggingface
-        # TODO: This needs to be fixed when new dlc is published
+        # TODO: Remove this logic whenever 0.20.0 image is out of service
         if (
             serving_properties["option.entryPoint"] == "djl_python.huggingface"
             and self.dtype
             and self.dtype != "auto"
+            and self.djl_version
+            and int(self.djl_version.split(".")[1]) < 21
         ):
             serving_properties["option.dtype"] = "auto"
             serving_properties.pop("option.load_in_8bit", None)

src/sagemaker/experiments/run.py

@@ -633,7 +633,10 @@ def _extract_run_name_from_tc_name(trial_component_name: str, experiment_name: s
         Returns:
             str: The name of the Run object supplied by a user.
         """
-        return trial_component_name.replace("{}{}".format(experiment_name, DELIMITER), "", 1)
+        # TODO: we should revert the lower casting once backend fix reaches prod
+        return trial_component_name.replace(
+            "{}{}".format(experiment_name.lower(), DELIMITER), "", 1
+        )
 
     @staticmethod
     def _append_run_tc_label_to_tags(tags: Optional[List[Dict[str, str]]] = None) -> list:
@@ -869,6 +872,8 @@ def list_runs(
     Returns:
         list: A list of ``Run`` objects.
     """
+
+    # all trial components retrieved by default
     tc_summaries = _TrialComponent.list(
         experiment_name=experiment_name,
         created_before=created_before,

src/sagemaker/inference_recommender/inference_recommender_mixin.py

@@ -145,10 +145,10 @@ def right_size(
         )
 
         if endpoint_configurations or traffic_pattern or stopping_conditions or resource_limit:
-            LOGGER.info("Advance Job parameters were specified. Running Advanced job...")
+            LOGGER.info("Advanced Job parameters were specified. Running Advanced job...")
             job_type = "Advanced"
         else:
-            LOGGER.info("Advance Job parameters were not specified. Running Default job...")
+            LOGGER.info("Advanced Job parameters were not specified. Running Default job...")
             job_type = "Default"
 
         self._init_sagemaker_session_if_does_not_exist()

tests/integ/sagemaker/experiments/test_run.py

@@ -642,6 +642,38 @@ def test_list(run_obj, sagemaker_session):
     assert run_tcs[0].experiment_config == run_obj.experiment_config
 
 
+def test_list_twice(run_obj, sagemaker_session):
+    tc1 = _TrialComponent.create(
+        trial_component_name=f"non-run-tc1-{name()}",
+        sagemaker_session=sagemaker_session,
+    )
+    tc2 = _TrialComponent.create(
+        trial_component_name=f"non-run-tc2-{name()}",
+        sagemaker_session=sagemaker_session,
+        tags=TAGS,
+    )
+    run_obj._trial.add_trial_component(tc1)
+    run_obj._trial.add_trial_component(tc2)
+
+    run_tcs = list_runs(
+        experiment_name=run_obj.experiment_name, sagemaker_session=sagemaker_session
+    )
+    assert len(run_tcs) == 1
+    assert run_tcs[0].run_name == run_obj.run_name
+    assert run_tcs[0].experiment_name == run_obj.experiment_name
+    assert run_tcs[0].experiment_config == run_obj.experiment_config
+
+    # note the experiment name used by run_obj is already mixed case and so
+    # covers the mixed case experiment name double create issue
+    run_tcs_second_result = list_runs(
+        experiment_name=run_obj.experiment_name, sagemaker_session=sagemaker_session
+    )
+    assert len(run_tcs) == 1
+    assert run_tcs_second_result[0].run_name == run_obj.run_name
+    assert run_tcs_second_result[0].experiment_name == run_obj.experiment_name
+    assert run_tcs_second_result[0].experiment_config == run_obj.experiment_config
+
+
 def _generate_estimator(
     exp_name,
     sdk_tar,

tests/integ/utils.py

@@ -19,12 +19,15 @@
 from tests.conftest import NO_P3_REGIONS, NO_M4_REGIONS
 from sagemaker.exceptions import CapacityError
 
+P2_INSTANCES = ["ml.p2.xlarge", "ml.p2.8xlarge", "ml.p2.16xlarge"]
+P3_INSTANCES = ["ml.p3.2xlarge"]
+
 
 def gpu_list(region):
     if region in NO_P3_REGIONS:
-        return ["ml.p2.xlarge"]
+        return P2_INSTANCES
     else:
-        return ["ml.p3.2xlarge", "ml.p2.xlarge"]
+        return [*P2_INSTANCES, *P3_INSTANCES]
 
 
 def cpu_list(region):

tests/unit/sagemaker/experiments/helpers.py

@@ -17,6 +17,7 @@
 
 
 TEST_EXP_NAME = "my-experiment"
+TEST_EXP_NAME_MIXED_CASE = "My-eXpeRiMeNt"
 TEST_RUN_NAME = "my-run"
 TEST_EXP_DISPLAY_NAME = "my-experiment-display-name"
 TEST_RUN_DISPLAY_NAME = "my-run-display-name"

tests/unit/sagemaker/experiments/test_run.py

@@ -48,6 +48,7 @@
     mock_trial_load_or_create_func,
     mock_tc_load_or_create_func,
     TEST_EXP_NAME,
+    TEST_EXP_NAME_MIXED_CASE,
     TEST_RUN_NAME,
     TEST_EXP_DISPLAY_NAME,
     TEST_RUN_DISPLAY_NAME,
@@ -779,7 +780,9 @@ def test_list(mock_tc_search, mock_tc_list, mock_tc_load, run_obj, sagemaker_ses
     ]
     mock_tc_list.return_value = [
         TrialComponentSummary(
-            trial_component_name=Run._generate_trial_component_name("A" + str(i), TEST_EXP_NAME),
+            trial_component_name=Run._generate_trial_component_name(
+                "A" + str(i), TEST_EXP_NAME_MIXED_CASE
+            ),
             trial_component_arn="b" + str(i),
             display_name="C" + str(i),
             source_arn="D" + str(i),
@@ -798,7 +801,7 @@ def test_list(mock_tc_search, mock_tc_list, mock_tc_load, run_obj, sagemaker_ses
         (
             _TrialComponent(
                 trial_component_name=Run._generate_trial_component_name(
-                    "a" + str(i), TEST_EXP_NAME
+                    "a" + str(i), TEST_EXP_NAME_MIXED_CASE
                 ),
                 trial_component_arn="b" + str(i),
                 display_name="C" + str(i),
@@ -818,14 +821,14 @@ def test_list(mock_tc_search, mock_tc_list, mock_tc_load, run_obj, sagemaker_ses
     ]
 
     run_list = list_runs(
-        experiment_name=TEST_EXP_NAME,
+        experiment_name=TEST_EXP_NAME_MIXED_CASE,
         sort_by=SortByType.CREATION_TIME,
         sort_order=SortOrderType.ASCENDING,
         sagemaker_session=sagemaker_session,
     )
 
     mock_tc_list.assert_called_once_with(
-        experiment_name=TEST_EXP_NAME,
+        experiment_name=TEST_EXP_NAME_MIXED_CASE,
         created_before=None,
         created_after=None,
         sort_by="CreationTime",

tests/unit/test_djl_inference.py

@@ -454,7 +454,7 @@ def test_generate_serving_properties_with_valid_configurations(
         "option.entryPoint": "djl_python.huggingface",
         "option.s3url": VALID_UNCOMPRESSED_MODEL_DATA,
         "option.tensor_parallel_degree": 1,
-        "option.dtype": "auto",
+        "option.dtype": "fp32",
         "option.device_id": 4,
         "option.device_map": "balanced",
     }