apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-10-05 16:48:03 +08:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

[BugFix] [CP] fix max streaming tokens invalid (#3798)

Browse Source 
* Update serving_chat.py

* Update serving_completion.py
This commit is contained in:
ltd0924
2025-09-02 21:03:36 +08:00
committed by GitHub
parent 1e19833ba5
commit cc5430e4c2
2 changed files with 2 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
fastdeploy/entrypoints/openai/serving_chat.py
View File

@@ -148,7 +148,7 @@ class OpenAIServingChat:
if request.max_streaming_response_tokens is not None
else (request.metadata or {}).get("max_streaming_response_tokens", 1)
) # dierctly passed & passed in metadata
max_streaming_response_tokens = max(1, max_streaming_response_tokens)
enable_thinking = request.chat_template_kwargs.get("enable_thinking") if request.chat_template_kwargs else None
if enable_thinking is None:
enable_thinking = request.metadata.get("enable_thinking") if request.metadata else None
@@ -316,11 +316,6 @@ class OpenAIServingChat:
api_server_logger.info(f"Chat Streaming response last send: {chunk.model_dump_json()}")
choices = []
if choices:
chunk.choices = choices
yield f"data: {chunk.model_dump_json(exclude_unset=True)}\n\n"
choices = []
if include_usage:
completion_tokens = previous_num_tokens
usage = UsageInfo(

5
fastdeploy/entrypoints/openai/serving_completion.py
View File

@@ -285,6 +285,7 @@ class OpenAIServingCompletion:
if request.max_streaming_response_tokens is not None
else (request.suffix or {}).get("max_streaming_response_tokens", 1)
) # dierctly passed & passed in suffix
max_streaming_response_tokens = max(max_streaming_response_tokens, 1)
choices = []
chunk = CompletionStreamResponse(
id=request_id,
@@ -418,10 +419,6 @@ class OpenAIServingCompletion:
)
yield f"data: {usage_chunk.model_dump_json(exclude_unset=True)}\n\n"
api_server_logger.info(f"Completion Streaming response last send: {chunk.model_dump_json()}")
if choices:
chunk.choices = choices
yield f"data: {chunk.model_dump_json(exclude_unset=True)}\n\n"
choices = []
except Exception as e:
yield f"data: {ErrorResponse(message=str(e), code=400).model_dump_json(exclude_unset=True)}\n\n"