Return usage for openai streaming requests (#1663)

Author: ichernev

vllm/entrypoints/openai/api_server.py

@@ -245,6 +245,7 @@ def create_stream_response_json(
         index: int,
         text: str,
         finish_reason: Optional[str] = None,
+        usage: Optional[UsageInfo] = None,
     ) -> str:
         choice_data = ChatCompletionResponseStreamChoice(
             index=index,
@@ -257,7 +258,10 @@ def create_stream_response_json(
             model=model_name,
             choices=[choice_data],
         )
-        response_json = response.json(ensure_ascii=False)
+        if usage is not None:
+            response.usage = usage
+        # exclude unset to leave details out of each sse
+        response_json = response.json(exclude_unset=True, ensure_ascii=False)
 
         return response_json
 
@@ -283,17 +287,25 @@ async def completion_stream_generator() -> AsyncGenerator[str, None]:
                 i = output.index
                 delta_text = output.text[len(previous_texts[i]):]
                 previous_texts[i] = output.text
-                previous_num_tokens[i] = len(output.token_ids)
+                completion_tokens = len(output.token_ids)
+                previous_num_tokens[i] = completion_tokens
                 response_json = create_stream_response_json(
                     index=i,
                     text=delta_text,
                 )
                 yield f"data: {response_json}\n\n"
                 if output.finish_reason is not None:
+                    prompt_tokens = len(res.prompt_token_ids)
+                    final_usage = UsageInfo(
+                        prompt_tokens=prompt_tokens,
+                        completion_tokens=completion_tokens,
+                        total_tokens=prompt_tokens + completion_tokens,
+                    )
                     response_json = create_stream_response_json(
                         index=i,
                         text="",
                         finish_reason=output.finish_reason,
+                        usage=final_usage,
                     )
                     yield f"data: {response_json}\n\n"
         yield "data: [DONE]\n\n"
@@ -462,6 +474,7 @@ def create_stream_response_json(
         text: str,
         logprobs: Optional[LogProbs] = None,
         finish_reason: Optional[str] = None,
+        usage: Optional[UsageInfo] = None,
     ) -> str:
         choice_data = CompletionResponseStreamChoice(
             index=index,
@@ -475,7 +488,9 @@ def create_stream_response_json(
             model=model_name,
             choices=[choice_data],
         )
-        response_json = response.json(ensure_ascii=False)
+        if usage is not None:
+            response.usage = usage
+        response_json = response.json(exclude_unset=True, ensure_ascii=False)
 
         return response_json
 
@@ -505,11 +520,19 @@ async def completion_stream_generator() -> AsyncGenerator[str, None]:
                 if output.finish_reason is not None:
                     logprobs = (LogProbs()
                                 if request.logprobs is not None else None)
+                    prompt_tokens = len(res.prompt_token_ids)
+                    completion_tokens = len(output.token_ids)
+                    final_usage = UsageInfo(
+                        prompt_tokens=prompt_tokens,
+                        completion_tokens=completion_tokens,
+                        total_tokens=prompt_tokens + completion_tokens,
+                    )
                     response_json = create_stream_response_json(
                         index=i,
                         text="",
                         logprobs=logprobs,
                         finish_reason=output.finish_reason,
+                        usage=final_usage,
                     )
                     yield f"data: {response_json}\n\n"
         yield "data: [DONE]\n\n"

vllm/entrypoints/openai/protocol.py

@@ -139,6 +139,7 @@ class CompletionStreamResponse(BaseModel):
     created: int = Field(default_factory=lambda: int(time.time()))
     model: str
     choices: List[CompletionResponseStreamChoice]
+    usage: Optional[UsageInfo]
 
 
 class ChatMessage(BaseModel):
@@ -178,3 +179,5 @@ class ChatCompletionStreamResponse(BaseModel):
     created: int = Field(default_factory=lambda: int(time.time()))
     model: str
     choices: List[ChatCompletionResponseStreamChoice]
+    usage: Optional[UsageInfo] = Field(
+        default=None, description="data about request and response")