Exception if remote server is not in explicit mode (#754)

* Throw exception if remote server is not in explicit mode

* Fix based on review

* Remove comment

docs/config_search.md

@@ -148,14 +148,6 @@ _This will perform an Automatic Brute Search with instance group counts: 3-5, ba
 
 ---
 
-### **Interaction with Remote Triton Launch Mode**
-
-When the triton launch mode is remote, _\*\*only concurrency or request rate values can be swept._\*\*<br>
-
-Model Analyzer will ignore any model config parameters because we have no way of accessing and modifying the model repository of the remote Triton Server.
-
----
-
 ## Manual Brute Search
 
 **Default brute search mode when any model config parameters or parameters are specified**

model_analyzer/triton/client/client.py

@@ -105,6 +105,10 @@ def load_model(self, model_name, variant_name="", config_str=None):
             return None
         except Exception as e:
             logger.info(f"Model {variant_name} load failed: {e}")
+            if "polling is enabled" in e.message():
+                raise TritonModelAnalyzerException(
+                    "The remote Tritonserver needs to be launched in EXPLICIT mode"
+                )
             return -1
 
     def unload_model(self, model_name):

tests/mocks/mock_client.py

@@ -16,6 +16,8 @@
 
 from unittest.mock import ANY, MagicMock, Mock, patch
 
+from tritonclient.http import InferenceServerException
+
 from .mock_base import MockBase
 
 
@@ -140,8 +142,12 @@ def raise_exception_on_load(self):
         InferenceServerException
         """
 
-        self.grpc_mock.return_value.load_model.side_effect = Exception
-        self.http_mock.return_value.load_model.side_effect = Exception
+        self.grpc_mock.return_value.load_model.side_effect = InferenceServerException(
+            ""
+        )
+        self.http_mock.return_value.load_model.side_effect = InferenceServerException(
+            ""
+        )
 
     def raise_exception_on_unload(self):
         """