fix: fix ml_inf to ml_inf1 in Neo multi-version support (aws#3405)

Author: HappyAmazonian

src/sagemaker/model.py

@@ -789,7 +789,7 @@ def multi_version_compilation_supported(
                 }
                 if target_instance_type in NEO_IOC_TARGET_DEVICES:
                     return framework in multi_version_frameworks_support_mapping["neo_ioc_targets"]
-                if target_instance_type == "ml_inf":
+                if target_instance_type == "ml_inf1":
                     return framework in multi_version_frameworks_support_mapping["inferentia"]
             return False
 

tests/unit/sagemaker/model/test_neo.py

@@ -47,7 +47,7 @@ def test_compile_model_for_inferentia(sagemaker_session):
     )
     model = _create_model(sagemaker_session)
     model.compile(
-        target_instance_family="ml_inf",
+        target_instance_family="ml_inf1",
         input_shape={"data": [1, 3, 1024, 1024]},
         output_path="s3://output",
         role="role",
@@ -313,7 +313,7 @@ def test_compile_with_pytorch_neo_in_ml_inf(session):
 
     model = _create_model()
     model.compile(
-        target_instance_family="ml_inf",
+        target_instance_family="ml_inf1",
         input_shape={"data": [1, 3, 1024, 1024]},
         output_path="s3://output",
         role="role",
@@ -336,7 +336,7 @@ def test_compile_with_tensorflow_neo_in_ml_inf(session):
 
     model = _create_model()
     model.compile(
-        target_instance_family="ml_inf",
+        target_instance_family="ml_inf1",
         input_shape={"data": [1, 3, 1024, 1024]},
         output_path="s3://output",
         role="role",