Directly use customer-provided endpoint name for ModelBuilder deployment.

src/sagemaker/serve/builder/model_builder.py

@@ -116,7 +116,7 @@
     validate_image_uri_and_hardware,
 )
 from sagemaker.serverless import ServerlessInferenceConfig
-from sagemaker.utils import Tags, unique_name_from_base
+from sagemaker.utils import Tags
 from sagemaker.workflow.entities import PipelineVariable
 from sagemaker.huggingface.llm_utils import (
     get_huggingface_model_metadata,
@@ -1983,8 +1983,6 @@ def deploy(
         """
         if not hasattr(self, "built_model") and not hasattr(self, "_deployables"):
             raise ValueError("Model needs to be built before deploying")
-        if not update_endpoint:
-            endpoint_name = unique_name_from_base(endpoint_name)
 
         if not hasattr(self, "_deployables"):
             if not inference_config:  # Real-time Deployment

tests/integ/sagemaker/serve/test_base_model_builder_deploy.py

@@ -185,7 +185,7 @@ def invoke(self, input_object: object, model: object):
 
 def test_real_time_deployment(xgboost_model_builder):
     real_time_predictor = xgboost_model_builder.deploy(
-        endpoint_name="test", initial_instance_count=1
+        endpoint_name=f"test-{uuid.uuid1().hex}", initial_instance_count=1
     )
 
     assert real_time_predictor is not None
@@ -198,7 +198,7 @@ def test_real_time_deployment(xgboost_model_builder):
 
 def test_serverless_deployment(xgboost_model_builder):
     serverless_predictor = xgboost_model_builder.deploy(
-        endpoint_name="test1", inference_config=ServerlessInferenceConfig()
+        endpoint_name=f"test1-{uuid.uuid1().hex}", inference_config=ServerlessInferenceConfig()
     )
 
     assert serverless_predictor is not None

tests/integ/sagemaker/serve/test_serve_model_builder_inference_component_happy.py

@@ -14,6 +14,7 @@
 
 import pytest
 import tests.integ
+import uuid
 
 from botocore.exceptions import ClientError
 from sagemaker.predictor import Predictor
@@ -88,7 +89,7 @@ def test_model_builder_ic_sagemaker_endpoint(
     with timeout(minutes=SERVE_SAGEMAKER_ENDPOINT_TIMEOUT):
         try:
             logger.info("Deploying and predicting in SAGEMAKER_ENDPOINT mode...")
-            endpoint_name = "llama-ic-endpoint-name"
+            endpoint_name = f"llama-ic-endpoint-name-{uuid.uuid1().hex}"
             predictors = chain.deploy(
                 instance_type=INSTANCE_TYPE,
                 initial_instance_count=1,

tests/unit/sagemaker/serve/builder/test_model_builder.py

@@ -4241,9 +4241,7 @@ def test_neuron_configurations_rule_set(self):
         "Batch",
     ],
 )
-@patch("sagemaker.serve.builder.model_builder.unique_name_from_base")
-def test_deploy(mock_unique_name_from_base, test_case):
-    mock_unique_name_from_base.return_value = "test"
+def test_deploy(test_case):
     model: Model = MagicMock()
     model_builder = ModelBuilder(
         model="meta-llama/Meta-Llama-3-8B-Instruct",