apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-10-04 16:22:57 +08:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

[Feature] Add return_token_ids, prompt_token_ids, and delete training, raw_request in request body (#2940)

Browse Source 
* [feat] add return_token_ids, prompt_token_ids, delete raw_request in request body

* [fix] return_token_ids not working in curl request

* [test] improve some test cases of return_token_ids and prompt_token_ids

* [fix] the server responds ok even if request.messages is an empty list
This commit is contained in:
李泳桦
2025-07-21 19:31:14 +08:00
committed by GitHub
parent 2845bde964
commit 8a619e9db5
8 changed files with 506 additions and 59 deletions
Download Patch File Download Diff File Expand all files Collapse all files

41
fastdeploy/entrypoints/openai/protocol.py
View File

@@ -124,6 +124,8 @@ class ChatMessage(BaseModel):
content: str
reasoning_content: Optional[str] = None
tool_calls: Optional[List[DeltaToolCall | ToolCall]] = None
prompt_token_ids: Optional[List[int]] = None
completion_token_ids: Optional[List[int]] = None
class ChatCompletionResponseChoice(BaseModel):
@@ -177,7 +179,8 @@ class DeltaMessage(BaseModel):
role: Optional[str] = None
content: Optional[str] = None
token_ids: Optional[List[int]] = None
prompt_token_ids: Optional[List[int]] = None
completion_token_ids: Optional[List[int]] = None
reasoning_content: Optional[str] = None
tool_calls: Optional[List[DeltaToolCall | ToolCall]] = None
@@ -214,7 +217,8 @@ class CompletionResponseChoice(BaseModel):
index: int
text: str
token_ids: Optional[List[int]] = None
prompt_token_ids: Optional[List[int]] = None
completion_token_ids: Optional[List[int]] = None
arrival_time: Optional[float] = None
logprobs: Optional[int] = None
reasoning_content: Optional[str] = None
@@ -243,7 +247,8 @@ class CompletionResponseStreamChoice(BaseModel):
index: int
text: str
arrival_time: float = None
token_ids: Optional[List[int]] = None
prompt_token_ids: Optional[List[int]] = None
completion_token_ids: Optional[List[int]] = None
logprobs: Optional[float] = None
reasoning_content: Optional[str] = None
finish_reason: Optional[Literal["stop", "length", "tool_calls"]] = None
@@ -341,6 +346,9 @@ class CompletionRequest(BaseModel):
top_k: Optional[int] = None
min_p: Optional[float] = None
user: Optional[str] = None
extra_body: Optional[dict] = None
return_token_ids: Optional[bool] = False
prompt_token_ids: Optional[List[int]] = None
response_format: Optional[AnyResponseFormat] = None
guided_json: Optional[Union[str, dict, BaseModel]] = None
@@ -373,9 +381,13 @@ class CompletionRequest(BaseModel):
if prompt is not None:
req_dict["prompt"] = prompt
if isinstance(prompt[0], int):
req_dict["prompt_token_ids"] = prompt
del req_dict["prompt"]
if self.prompt_token_ids is not None or \
(self.extra_body is not None and self.extra_body.get("prompt_token_ids") is not None):
req_dict["prompt_token_ids"] = self.prompt_token_ids
if "prompt" in req_dict:
del req_dict["prompt"]
else:
assert len(prompt) > 0
guided_json_object = None
if self.response_format is not None:
@@ -464,6 +476,9 @@ class ChatCompletionRequest(BaseModel):
min_p: Optional[float] = None
user: Optional[str] = None
metadata: Optional[dict] = None
extra_body: Optional[dict] = None
return_token_ids: Optional[bool] = False
prompt_token_ids: Optional[List[int]] = None
response_format: Optional[AnyResponseFormat] = None
guided_json: Optional[Union[str, dict, BaseModel]] = None
@@ -499,12 +514,14 @@ class ChatCompletionRequest(BaseModel):
for key, value in self.dict().items():
if value is not None:
req_dict[key] = value
if isinstance(self.messages[0], int):
req_dict["prompt_token_ids"] = self.messages
del req_dict["messages"]
if "raw_request" in req_dict and not req_dict["raw_request"]:
req_dict["prompt"] = req_dict["messages"][0]["content"]
del req_dict["messages"]
if self.prompt_token_ids is not None or \
(self.extra_body is not None and self.extra_body.get("prompt_token_ids") is not None):
req_dict["prompt_token_ids"] = self.prompt_token_ids
if "messages" in req_dict:
del req_dict["messages"]
else:
assert len(self.messages) > 0
guided_json_object = None
if self.response_format is not None: