RYDE-WORK/ktransformers
1
0
Fork 0
mirror of https://github.com/RYDE-WORK/ktransformers.git synced 2026-02-08 16:08:00 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity

Also /chat/completions

Browse Source
This commit is contained in:
lazymio 2025-02-24 21:08:36 +08:00
parent bf36547f98
commit 05ad288453
No known key found for this signature in database
GPG Key ID: DFF27E34A47CB873
2 changed files with 5 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
ktransformers/server/api/openai/endpoints/chat.py
View File

@ -28,13 +28,13 @@ async def chat_completion(request:Request,create:ChatCompletionCreate):
if create.stream: if create.stream:
async def inner(): async def inner():
chunk = ChatCompletionChunk(id=id,object='chat.completion.chunk',created=int(time())) chunk = ChatCompletionChunk(id=id,object='chat.completion.chunk',created=int(time()))
async for token in interface.inference(input_message,id): async for token in interface.inference(input_message,id,create.temperature,create.top_p,create.repetition_penalty):
chunk.set_token(token) chunk.set_token(token)
yield chunk yield chunk
return chat_stream_response(request,inner()) return chat_stream_response(request,inner())
else: else:
comp = ChatCompletionObject(id=id,object='chat.completion',created=int(time())) comp = ChatCompletionObject(id=id,object='chat.completion',created=int(time()))
comp.usage = Usage(completion_tokens=1, prompt_tokens=1, total_tokens=2) comp.usage = Usage(completion_tokens=1, prompt_tokens=1, total_tokens=2)
async for token in interface.inference(input_message,id): async for token in interface.inference(input_message,id,create.temperature,create.top_p,create.repetition_penalty):
comp.append_token(token) comp.append_token(token)
return comp return comp

3
ktransformers/server/schemas/endpoints/chat.py
View File

@ -25,6 +25,9 @@ class ChatCompletionCreate(BaseModel):
messages: List[Message] messages: List[Message]
model : str model : str
stream : bool = False stream : bool = False
temperature: Optional[float]
top_p: Optional[float]
repetition_penalty: Optional[float]
def get_tokenizer_messages(self): def get_tokenizer_messages(self):
return [m.to_tokenizer_message() for m in self.messages] return [m.to_tokenizer_message() for m in self.messages]