(proxy - RPS) - Get 2K RPS at 4 instances, minor fix aiohttp_openai/ (

#7659)

* speed up transform_response

* use 2 workers

* undo changes to uvicorn

* ci/cd run again

litellm/llms/aiohttp_openai/chat/transformation.py

@@ -13,7 +13,7 @@
 
 from litellm.llms.openai_like.chat.transformation import OpenAILikeChatConfig
 from litellm.types.llms.openai import AllMessageValues
-from litellm.types.utils import ModelResponse
+from litellm.types.utils import Choices, ModelResponse
 
 if TYPE_CHECKING:
     from litellm.litellm_core_utils.litellm_logging import Logging as _LiteLLMLoggingObj
@@ -66,4 +66,12 @@ async def transform_response(  # type: ignore
         json_mode: Optional[bool] = None,
     ) -> ModelResponse:
         _json_response = await raw_response.json()
-        return ModelResponse(**_json_response)
+        model_response.id = _json_response.get("id")
+        model_response.choices = [
+            Choices(**choice) for choice in _json_response.get("choices")
+        ]
+        model_response.created = _json_response.get("created")
+        model_response.model = _json_response.get("model")
+        model_response.object = _json_response.get("object")
+        model_response.system_fingerprint = _json_response.get("system_fingerprint")
+        return model_response

tests/llm_translation/test_aiohttp_openai.py

@@ -6,7 +6,7 @@
 
 sys.path.insert(
     0, os.path.abspath("../../")
-)  # Adds the parent directory to the system path
+)  # Adds the parent directory to the system-path
 
 import litellm