Morpheus tests (#1633)

* Parameterized intelligent defaults tests

* Parameterized intelligent defaults tests

* Parameterized intelligent defaults tests

* Tests for all Model Builder deployment modes

* Fix

* CodeStyle Fixes

* CodeStyle Fixes

* Add Deepdiff dependency

* Add Deepdiff dependency

* Add Codestyle fix

Author: nargokul

requirements/extras/test_requirements.txt

@@ -49,3 +49,4 @@ uvicorn>=0.30.1
 fastapi==0.115.4
 nest-asyncio
 sagemaker-mlflow>=0.1.0
+deepdiff>=8.0.0

src/sagemaker/modules/train/model_trainer.py

@@ -235,7 +235,6 @@ class ModelTrainer(BaseModel):
         "role",
         "base_job_name",
         "source_code",
-        "distributed",
         "compute",
         "networking",
         "stopping_condition",
@@ -251,7 +250,6 @@ class ModelTrainer(BaseModel):
 
     SERIALIZABLE_CONFIG_ATTRIBUTES: ClassVar[Any] = {
         "source_code": SourceCode,
-        "distributed": DistributedConfig,
         "compute": Compute,
         "networking": Networking,
         "stopping_condition": StoppingCondition,

src/sagemaker/serve/builder/model_builder.py

@@ -35,7 +35,7 @@
 from sagemaker.jumpstart.accessors import JumpStartS3PayloadAccessor
 from sagemaker.jumpstart.utils import get_jumpstart_content_bucket
 from sagemaker.s3 import S3Downloader
-from sagemaker import Session, utils
+from sagemaker import Session
 from sagemaker.model import Model
 from sagemaker.base_predictor import PredictorBase
 from sagemaker.serializers import NumpySerializer, TorchTensorSerializer
@@ -112,7 +112,7 @@
     validate_image_uri_and_hardware,
 )
 from sagemaker.serverless import ServerlessInferenceConfig
-from sagemaker.utils import Tags
+from sagemaker.utils import Tags, unique_name_from_base
 from sagemaker.workflow.entities import PipelineVariable
 from sagemaker.huggingface.llm_utils import (
     get_huggingface_model_metadata,
@@ -1621,7 +1621,7 @@ def deploy(
         """
         if not hasattr(self, "built_model"):
             raise ValueError("Model Needs to be built before deploying")
-        endpoint_name = utils.unique_name_from_base(endpoint_name)
+        endpoint_name = unique_name_from_base(endpoint_name)
         if not inference_config:  # Real-time Deployment
             return self.built_model.deploy(
                 instance_type=self.instance_type,

tests/unit/sagemaker/modules/train/test_model_trainer.py

@@ -194,17 +194,57 @@ def test_train_with_default_params(mock_training_job, model_trainer):
     training_job_instance.wait.assert_called_once_with(logs=True)
 
 
+@pytest.mark.parametrize(
+    "default_config",
+    [
+        {
+            "path_name": "sourceCode",
+            "config_value": {"command": "echo 'Hello World' && env"},
+        },
+        {
+            "path_name": "compute",
+            "config_value": {"volume_size_in_gb": 45},
+        },
+        {
+            "path_name": "networking",
+            "config_value": {
+                "enable_network_isolation": True,
+                "security_group_ids": ["sg-1"],
+                "subnets": ["subnet-1"],
+            },
+        },
+        {
+            "path_name": "stoppingCondition",
+            "config_value": {"max_runtime_in_seconds": 15},
+        },
+        {
+            "path_name": "trainingImageConfig",
+            "config_value": {"training_repository_access_mode": "private"},
+        },
+        {
+            "path_name": "outputDataConfig",
+            "config_value": {"s3_output_path": "Sample S3 path"},
+        },
+        {
+            "path_name": "checkpointConfig",
+            "config_value": {"s3_uri": "sample uri"},
+        },
+    ],
+)
 @patch("sagemaker.modules.train.model_trainer.TrainingJob")
 @patch("sagemaker.modules.train.model_trainer.resolve_value_from_config")
 @patch("sagemaker.modules.train.model_trainer.ModelTrainer.create_input_data_channel")
 def test_train_with_intelligent_defaults(
-    mock_create_input_data_channel, mock_resolve_value_from_config, mock_training_job, model_trainer
+    mock_create_input_data_channel,
+    mock_resolve_value_from_config,
+    mock_training_job,
+    default_config,
+    model_trainer,
 ):
-    source_code_path = _simple_path(SAGEMAKER, PYTHON_SDK, MODULES, MODEL_TRAINER, "sourceCode")
-
     mock_resolve_value_from_config.side_effect = lambda **kwargs: (
-        {"command": "echo 'Hello World' && env"}
-        if kwargs["config_path"] == source_code_path
+        default_config["config_value"]
+        if kwargs["config_path"]
+        == _simple_path(SAGEMAKER, PYTHON_SDK, MODULES, MODEL_TRAINER, default_config["path_name"])
         else None
     )