[server] Fix tests + allow server to parse generation kwargs (#1311)

* update predict to call the pipeline directly

* fix tests

src/deepsparse/server/sagemaker.py

@@ -57,11 +57,25 @@ def _add_inference_endpoints(
 
         if hasattr(pipeline.input_schema, "from_files"):
             routes_and_fns.append(
-                (route, partial(Server.predict_from_files, ProxyPipeline(pipeline)))
+                (
+                    route + "/from_files",
+                    partial(
+                        Server.predict_from_files,
+                        ProxyPipeline(pipeline),
+                        self.server_config.system_logging,
+                    ),
+                )
             )
         else:
             routes_and_fns.append(
-                (route, partial(Server.predict, ProxyPipeline(pipeline)))
+                (
+                    route,
+                    partial(
+                        Server.predict,
+                        ProxyPipeline(pipeline),
+                        self.server_config.system_logging,
+                    ),
+                )
             )
 
         self._update_routes(

src/deepsparse/server/server.py

@@ -246,19 +246,27 @@ async def predict(
         system_logging_config: SystemLoggingConfig,
         raw_request: Request,
     ):
-        request = proxy_pipeline.pipeline.input_schema(**await raw_request.json())
-        pipeline_outputs = proxy_pipeline.pipeline(request)
+        pipeline_outputs = proxy_pipeline.pipeline(**await raw_request.json())
         server_logger = proxy_pipeline.pipeline.logger
         if server_logger:
             log_system_information(
                 server_logger=server_logger, system_logging_config=system_logging_config
             )
-        pipeline_outputs = prep_outputs_for_serialization(pipeline_outputs)
-        return pipeline_outputs
+        return prep_outputs_for_serialization(pipeline_outputs)
 
     @staticmethod
-    def predict_from_files(proxy_pipeline: ProxyPipeline, request: List[UploadFile]):
+    def predict_from_files(
+        proxy_pipeline: ProxyPipeline,
+        system_logging_config: SystemLoggingConfig,
+        request: List[UploadFile],
+    ):
         request = proxy_pipeline.pipeline.input_schema.from_files(
             (file.file for file in request), from_server=True
         )
-        return Server.predict(request)
+        pipeline_outputs = proxy_pipeline.pipeline(request)
+        server_logger = proxy_pipeline.pipeline.logger
+        if server_logger:
+            log_system_information(
+                server_logger=server_logger, system_logging_config=system_logging_config
+            )
+        return prep_outputs_for_serialization(pipeline_outputs)

tests/server/test_endpoints.py

@@ -38,8 +38,8 @@ class StrSchema(BaseModel):
     value: str
 
 
-def parse(v: StrSchema) -> int:
-    return int(v.value)
+def parse(value) -> int:
+    return int(value)
 
 
 class TestStatusEndpoints:
@@ -106,7 +106,7 @@ def test_add_model_endpoint(
     ):
         mock_pipeline = Mock(
             side_effect=parse,
-            input_schema=StrSchema,
+            input_schema=str,
             output_schema=int,
             logger=MultiLogger([]),
         )
@@ -146,6 +146,7 @@ def test_add_model_endpoint_with_from_files(self, server, app):
         assert app.routes[-1].path == "/v2/models/predict/parse_int/infer/from_files"
         assert app.routes[-1].endpoint.func.__annotations__ == {
             "proxy_pipeline": ProxyPipeline,
+            "system_logging_config": SystemLoggingConfig,
             "request": List[UploadFile],
         }
         assert app.routes[-1].response_model is int
@@ -159,9 +160,12 @@ def test_sagemaker_only_adds_one_endpoint(self, sagemaker_server, app):
             pipeline=Mock(input_schema=FromFilesSchema, output_schema=int),
         )
         assert len(app.routes) == num_routes + 1
-        assert app.routes[-1].path == "/invocations/predict/parse_int/infer"
+        num_routes = len(app.routes)
+
+        assert app.routes[-1].path == "/invocations/predict/parse_int/infer/from_files"
         assert app.routes[-1].endpoint.func.__annotations__ == {
             "proxy_pipeline": ProxyPipeline,
+            "system_logging_config": SystemLoggingConfig,
             "request": List[UploadFile],
         }
 
@@ -174,7 +178,8 @@ def test_sagemaker_only_adds_one_endpoint(self, sagemaker_server, app):
         assert app.routes[-1].path == "/invocations/predict/parse_int/infer"
         assert app.routes[-1].endpoint.func.__annotations__ == {
             "proxy_pipeline": ProxyPipeline,
-            "request": List[UploadFile],
+            "system_logging_config": SystemLoggingConfig,
+            "raw_request": Request,
         }
 
     def test_add_endpoint_with_no_route_specified(self, server, app):