[Frontend] Adjust try/except blocks in API impl (#10056)

Signed-off-by: Nick Hill <nhill@redhat.com>

vllm/entrypoints/openai/serving_completion.py

@@ -189,13 +189,7 @@ class OpenAIServingCompletion(OpenAIServing):
         try:
             async for i, res in result_generator:
                 final_res_batch[i] = res
-        except asyncio.CancelledError:
-            return self.create_error_response("Client disconnected")
-        except ValueError as e:
-            # TODO: Use a vllm-specific Validation Error
-            return self.create_error_response(str(e))
 
-        try:
             for i, final_res in enumerate(final_res_batch):
                 assert final_res is not None
 
@@ -217,6 +211,8 @@ class OpenAIServingCompletion(OpenAIServing):
                 tokenizer,
                 request_metadata,
             )
+        except asyncio.CancelledError:
+            return self.create_error_response("Client disconnected")
         except ValueError as e:
             # TODO: Use a vllm-specific Validation Error
             return self.create_error_response(str(e))

vllm/entrypoints/openai/serving_embedding.py

@@ -205,12 +205,8 @@ class OpenAIServingEmbedding(OpenAIServing):
         try:
             async for i, res in result_generator:
                 final_res_batch[i] = res
-        except asyncio.CancelledError:
-            return self.create_error_response("Client disconnected")
 
-        try:
-            for final_res in final_res_batch:
-                assert final_res is not None
+            assert all(final_res is not None for final_res in final_res_batch)
 
             final_res_batch_checked = cast(List[EmbeddingRequestOutput],
                                            final_res_batch)
@@ -218,6 +214,8 @@ class OpenAIServingEmbedding(OpenAIServing):
             response = request_output_to_embedding_response(
                 final_res_batch_checked, request_id, created_time, model_name,
                 encoding_format)
+        except asyncio.CancelledError:
+            return self.create_error_response("Client disconnected")
         except ValueError as e:
             # TODO: Use a vllm-specific Validation Error
             return self.create_error_response(str(e))