Skip to content

[Server] Support openai prefix cache #2515

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

Closed
esmeetu wants to merge 6 commits into from
Closed

[Server] Support openai prefix cache #2515

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
815e190
support openai server prefix pos
esmeetu Jan 20, 2024
4db6cd4
format
esmeetu Jan 20, 2024
897de6a
add prefix_stop str
esmeetu Jan 21, 2024
a7b90d7
fix
esmeetu Jan 28, 2024
3ff80ca
Merge remote-tracking branch 'upstream/main' into openai-prefix-cache
esmeetu Jan 28, 2024
6f9f0d2
format
esmeetu Jan 28, 2024
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
4 changes: 4 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 @@ -78,6 +78,8 @@ class ChatCompletionRequest(BaseModel):
echo: Optional[bool] = False
repetition_penalty: Optional[float] = 1.0
min_p: Optional[float] = 0.0
prefix_pos: Optional[int] = None
prefix_stop: Optional[str] = None
include_stop_str_in_output: Optional[bool] = False
length_penalty: Optional[float] = 1.0

Expand Down Expand Up @@ -131,6 +133,8 @@ class CompletionRequest(BaseModel):
spaces_between_special_tokens: Optional[bool] = True
repetition_penalty: Optional[float] = 1.0
min_p: Optional[float] = 0.0
prefix_pos: Optional[int] = None
prefix_stop: Optional[str] = None
include_stop_str_in_output: Optional[bool] = False
length_penalty: Optional[float] = 1.0

Expand Down
10 changes: 9 additions & 1 deletion vllm/entrypoints/openai/serving_chat.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -67,8 +67,16 @@ async def create_chat_completion(
except ValueError as e:
return self.create_error_response(str(e))

prefix_pos = request.prefix_pos
if request.prefix_stop is not None and request.prefix_stop in prompt:
prefix_index = prompt.index(request.prefix_stop)
prefix_pos = len(self.tokenizer.encode(prompt[:prefix_index])) - 1
prompt = prompt.replace(request.prefix_stop, '')

result_generator = self.engine.generate(prompt, sampling_params,
request_id, token_ids)
request_id, token_ids,
prefix_pos)

# Streaming response
if request.stream:
return self.chat_completion_stream_generator(
Expand Down
11 changes: 10 additions & 1 deletion vllm/entrypoints/openai/serving_completion.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -286,19 +286,28 @@ async def create_completion(self, request: CompletionRequest,
sampling_params = request.to_sampling_params()
prompt_is_tokens, prompts = parse_prompt_format(request.prompt)

prefix_pos = request.prefix_pos
for i, prompt in enumerate(prompts):
if prompt_is_tokens:
input_ids = self._validate_prompt_and_tokenize(
request, prompt_ids=prompt)
else:
# Parse prefix position by prompt and prefix_stop indicator
if request.prefix_stop is not None and request.prefix_stop in prompt:
prefix_index = prompt.index(request.prefix_stop)
prefix_pos = len(
self.tokenizer.encode(prompt[:prefix_index])) - 1
prompt = prompt.replace(request.prefix_stop, '')

input_ids = self._validate_prompt_and_tokenize(
request, prompt=prompt)

generators.append(
self.engine.generate(None,
sampling_params,
f"{request_id}-{i}",
prompt_token_ids=input_ids))
prompt_token_ids=input_ids,
prefix_pos=prefix_pos))
except ValueError as e:
return self.create_error_response(str(e))

Expand Down