Stop throwing exception on python binding when multiple EP available (#…

…17659)

Stop throwing the exception when the provider list is empty but there
are multiple available EPs.
Other language bindings throw no exception at all, this change will
align them up.

---------

Co-authored-by: Randy Shuai <rashuai@microsoft.com>

onnxruntime/python/onnxruntime_inference_collection.py

@@ -438,7 +438,7 @@ def _create_inference_session(self, providers, provider_options, disabled_optimi
 
         # Tensorrt can fall back to CUDA if it's explicitly assigned. All others fall back to CPU.
         if "TensorrtExecutionProvider" in available_providers:
-            if any(
+            if providers and any(
                 provider == "CUDAExecutionProvider"
                 or (isinstance(provider, tuple) and provider[0] == "CUDAExecutionProvider")
                 for provider in providers
@@ -448,7 +448,7 @@ def _create_inference_session(self, providers, provider_options, disabled_optimi
                 self._fallback_providers = ["CPUExecutionProvider"]
         # MIGraphX can fall back to ROCM if it's explicitly assigned. All others fall back to CPU.
         elif "MIGraphXExecutionProvider" in available_providers:
-            if any(
+            if providers and any(
                 provider == "ROCMExecutionProvider"
                 or (isinstance(provider, tuple) and provider[0] == "ROCMExecutionProvider")
                 for provider in providers
@@ -463,14 +463,6 @@ def _create_inference_session(self, providers, provider_options, disabled_optimi
         providers, provider_options = check_and_normalize_provider_args(
             providers, provider_options, available_providers
         )
-        if not providers and len(available_providers) > 1:
-            self.disable_fallback()
-            raise ValueError(
-                f"This ORT build has {available_providers} enabled. "
-                "Since ORT 1.9, you are required to explicitly set "
-                "the providers parameter when instantiating InferenceSession. For example, "
-                f"onnxruntime.InferenceSession(..., providers={available_providers}, ...)"
-            )
 
         session_options = self._sess_options if self._sess_options else C.get_default_session_options()
         if self._model_path:

onnxruntime/test/python/onnxruntime_test_python.py

@@ -80,11 +80,7 @@ def test_model_serialization(self):
             so.log_severity_level = 1
             so.logid = "TestModelSerialization"
             so.optimized_model_filepath = "./PythonApiTestOptimizedModel.onnx"
-            onnxrt.InferenceSession(
-                get_name("mul_1.onnx"),
-                sess_options=so,
-                providers=["CPUExecutionProvider"],
-            )
+            onnxrt.InferenceSession(get_name("mul_1.onnx"), sess_options=so)
             self.assertTrue(os.path.isfile(so.optimized_model_filepath))
             os.remove(so.optimized_model_filepath)
         except Fail as onnxruntime_error:
@@ -107,11 +103,7 @@ def test_model_serialization_with_external_initializers(self):
                 "session.optimized_model_external_initializers_file_name", external_initializers_file
             )
             so.add_session_config_entry("session.optimized_model_external_initializers_min_size_in_bytes", "100")
-            onnxrt.InferenceSession(
-                get_name("mnist.onnx"),
-                sess_options=so,
-                providers=["CPUExecutionProvider"],
-            )
+            onnxrt.InferenceSession(get_name("mnist.onnx"), sess_options=so)
             self.assertTrue(os.path.isfile(so.optimized_model_filepath))
             self.assertTrue(os.path.isfile(external_initializers_file))
             os.remove(so.optimized_model_filepath)
@@ -137,7 +129,7 @@ def test_model_serialization_with_external_initializers_to_directory(self):
                 "session.optimized_model_external_initializers_file_name", external_initializers_file
             )
             so.add_session_config_entry("session.optimized_model_external_initializers_min_size_in_bytes", "100")
-            onnxrt.InferenceSession(get_name("mnist.onnx"), sess_options=so, providers=["CPUExecutionProvider"])
+            onnxrt.InferenceSession(get_name("mnist.onnx"), sess_options=so)
             self.assertTrue(os.path.isfile(so.optimized_model_filepath))
             self.assertTrue(os.path.isfile(os.path.join(directory, external_initializers_file)))
             os.remove(so.optimized_model_filepath)
@@ -163,9 +155,7 @@ def test_model_serialization_with_original_external_initializers_to_directory(se
                 "session.optimized_model_external_initializers_file_name", external_initializers_file
             )
             so.add_session_config_entry("session.optimized_model_external_initializers_min_size_in_bytes", "100")
-            onnxrt.InferenceSession(
-                get_name("model_with_orig_ext_data.onnx"), sess_options=so, providers=["CPUExecutionProvider"]
-            )
+            onnxrt.InferenceSession(get_name("model_with_orig_ext_data.onnx"), sess_options=so)
             self.assertTrue(os.path.isfile(so.optimized_model_filepath))
             self.assertTrue(os.path.isfile(os.path.join(directory, external_initializers_file)))
             os.remove(so.optimized_model_filepath)
@@ -198,9 +188,7 @@ def test_model_serialization_with_original_external_initializers_to_current_dire
         # still refers to the original external data file. We shall fix this issue so that the
         # optimized model only refers to one external data file.
         so.add_session_config_entry("session.optimized_model_external_initializers_min_size_in_bytes", "10")
-        session1 = onnxrt.InferenceSession(
-            get_name("model_with_orig_ext_data.onnx"), sess_options=so, providers=["CPUExecutionProvider"]
-        )
+        session1 = onnxrt.InferenceSession(get_name("model_with_orig_ext_data.onnx"), sess_options=so)
         del session1
         self.assertTrue(os.path.isfile(optimized_model_filepath))
         self.assertTrue(os.path.isfile(external_initializers_file))
@@ -216,9 +204,7 @@ def test_model_serialization_with_original_external_initializers_to_current_dire
 
         # verify that we can load the optimized model with external data in current directory and save
         # optimized model with external data to current directory.
-        session2 = onnxrt.InferenceSession(
-            optimized_model_filepath, sess_options=so2, providers=["CPUExecutionProvider"]
-        )
+        session2 = onnxrt.InferenceSession(optimized_model_filepath, sess_options=so2)
         del session2
         self.assertTrue(os.path.isfile(optimized_model_filepath_2))
         self.assertTrue(os.path.isfile(external_initializers_file_2))
@@ -227,9 +213,7 @@ def test_model_serialization_with_original_external_initializers_to_current_dire
         os.remove(optimized_model_filepath)
         os.remove(external_initializers_file)
 
-        session3 = onnxrt.InferenceSession(
-            optimized_model_filepath_2, sess_options=onnxrt.SessionOptions(), providers=["CPUExecutionProvider"]
-        )
+        session3 = onnxrt.InferenceSession(optimized_model_filepath_2, sess_options=onnxrt.SessionOptions())
         del session3
 
         os.remove(optimized_model_filepath_2)