Skip to content

correct max_tokens by context_length instead of raise exception #2544

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

Merged
merrymercy merged 1 commit into from
Oct 12, 2023
Merged

correct max_tokens by context_length instead of raise exception #2544

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
35 changes: 11 additions & 24 deletions fastchat/serve/openai_api_server.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -144,6 +144,11 @@ async def check_model(request) -> Optional[JSONResponse]:


async def check_length(request, prompt, max_tokens, worker_addr):
if (
not isinstance(max_tokens, int) or max_tokens <= 0
): # model worker not support max_tokens=None
max_tokens = 1024 * 1024

context_len = await fetch_remote(
worker_addr + "/model_details", {"model": request.model}, "context_length"
)
Expand All @@ -152,17 +157,7 @@ async def check_length(request, prompt, max_tokens, worker_addr):
{"model": request.model, "prompt": prompt},
"count",
)
if token_num + max_tokens > context_len:
return create_error_response(
ErrorCode.CONTEXT_OVERFLOW,
f"This model's maximum context length is {context_len} tokens. "
f"However, you requested {max_tokens + token_num} tokens "
f"({token_num} in the messages, "
f"{max_tokens} in the completion). "
f"Please reduce the length of the messages or completion.",
)
else:
return None
return min(max_tokens, context_len - token_num)


def check_requests(request) -> Optional[JSONResponse]:
Expand Down Expand Up @@ -275,8 +270,6 @@ async def get_gen_params(
conv.append_message(conv.roles[1], None)
prompt = conv.get_prompt()

if max_tokens is None:
max_tokens = 512
gen_params = {
"model": model_name,
"prompt": prompt,
Expand Down Expand Up @@ -363,14 +356,12 @@ async def create_chat_completion(request: ChatCompletionRequest):
echo=False,
stop=request.stop,
)
error_check_ret = await check_length(
gen_params["max_new_tokens"] = await check_length(
request,
gen_params["prompt"],
gen_params["max_new_tokens"],
worker_addr,
)
if error_check_ret is not None:
return error_check_ret

if request.stream:
generator = chat_completion_stream_generator(
Expand Down Expand Up @@ -475,11 +466,9 @@ async def create_completion(request: CompletionRequest):

worker_addr = await get_worker_address(request.model)
for text in request.prompt:
error_check_ret = await check_length(
request, text, request.max_tokens, worker_addr
)
if error_check_ret is not None:
return error_check_ret
max_tokens = await check_length(request, text, request.max_tokens, worker_addr)
if isinstance(max_tokens, int) and max_tokens < request.max_tokens:
request.max_tokens = max_tokens

if request.stream:
generator = generate_completion_stream_generator(
Expand Down Expand Up @@ -734,14 +723,12 @@ async def create_chat_completion(request: APIChatCompletionRequest):
if request.repetition_penalty is not None:
gen_params["repetition_penalty"] = request.repetition_penalty

error_check_ret = await check_length(
gen_params["max_new_tokens"] = await check_length(
request,
gen_params["prompt"],
gen_params["max_new_tokens"],
worker_addr,
)
if error_check_ret is not None:
return error_check_ret

if request.stream:
generator = chat_completion_stream_generator(
Expand Down