Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Return usage for openai requests #1663

Merged
merged 1 commit into from
Nov 16, 2023
Merged

Return usage for openai requests #1663

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
29 changes: 26 additions & 3 deletions vllm/entrypoints/openai/api_server.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -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,
Expand All @@ -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

Expand All @@ -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"
Expand Down Expand Up @@ -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,
Expand All @@ -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

Expand Down Expand Up @@ -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"
Expand Down
3 changes: 3 additions & 0 deletions vllm/entrypoints/openai/protocol.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -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):
Expand Down Expand Up @@ -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")
Loading