xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-01-17 14:14:27 +08:00
Code Issues Packages Projects Releases Wiki Activity

[responsesAPI]add extra body parameters (#30532)

Browse Source 
Signed-off-by: Ri0S <aa248424@gmail.com>
This commit is contained in:
RioS 2025-12-15 04:25:45 +09:00 committed by GitHub
parent ae2e503dda
commit 9ccbf6b692
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 9 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
vllm/entrypoints/openai/protocol.py
View File

@ -320,6 +320,7 @@ class ResponsesRequest(OpenAIBaseModel):
max_tool_calls: int | None = None
metadata: Metadata | None = None
model: str | None = None
logit_bias: dict[str, float] | None = None
parallel_tool_calls: bool | None = True
previous_response_id: str | None = None
prompt: ResponsePrompt | None = None
@ -333,6 +334,7 @@ class ResponsesRequest(OpenAIBaseModel):
tools: list[Tool] = Field(default_factory=list)
top_logprobs: int | None = 0
top_p: float | None = None
top_k: int | None = None
truncation: Literal["auto", "disabled"] | None = "disabled"
user: str | None = None
@ -387,6 +389,7 @@ class ResponsesRequest(OpenAIBaseModel):
_DEFAULT_SAMPLING_PARAMS = {
"temperature": 1.0,
"top_p": 1.0,
"top_k": 0,
}
def to_sampling_params(
@ -408,6 +411,10 @@ class ResponsesRequest(OpenAIBaseModel):
top_p = default_sampling_params.get(
"top_p", self._DEFAULT_SAMPLING_PARAMS["top_p"]
)
if (top_k := self.top_k) is None:
top_k = default_sampling_params.get(
"top_k", self._DEFAULT_SAMPLING_PARAMS["top_k"]
)
stop_token_ids = default_sampling_params.get("stop_token_ids")
# Structured output
@ -428,6 +435,7 @@ class ResponsesRequest(OpenAIBaseModel):
return SamplingParams.from_optional(
temperature=temperature,
top_p=top_p,
top_k=top_k,
max_tokens=max_tokens,
logprobs=self.top_logprobs if self.is_include_output_logprobs() else None,
stop_token_ids=stop_token_ids,
@ -435,6 +443,7 @@ class ResponsesRequest(OpenAIBaseModel):
RequestOutputKind.DELTA if self.stream else RequestOutputKind.FINAL_ONLY
),
structured_outputs=structured_outputs,
logit_bias=self.logit_bias,
)
def is_include_output_logprobs(self) -> bool: