[gpt-oss] Support chat completion api (#22342)
This commit is contained in:
@@ -323,6 +323,7 @@ class ResponsesRequest(OpenAIBaseModel):
|
||||
if (top_p := self.top_p) is None:
|
||||
top_p = default_sampling_params.get(
|
||||
"top_p", self._DEFAULT_SAMPLING_PARAMS["top_p"])
|
||||
stop_token_ids = default_sampling_params.get("stop_token_ids")
|
||||
|
||||
# Structured output
|
||||
guided_decoding = None
|
||||
@@ -340,6 +341,7 @@ class ResponsesRequest(OpenAIBaseModel):
|
||||
top_p=top_p,
|
||||
max_tokens=max_tokens,
|
||||
logprobs=self.top_logprobs,
|
||||
stop_token_ids=stop_token_ids,
|
||||
output_kind=(RequestOutputKind.DELTA
|
||||
if self.stream else RequestOutputKind.FINAL_ONLY),
|
||||
guided_decoding=guided_decoding,
|
||||
@@ -404,6 +406,8 @@ class ChatCompletionRequest(OpenAIBaseModel):
|
||||
Literal["required"],
|
||||
ChatCompletionNamedToolChoiceParam,
|
||||
]] = "none"
|
||||
reasoning_effort: Optional[Literal["low", "medium", "high"]] = None
|
||||
include_reasoning: bool = True
|
||||
|
||||
# NOTE this will be ignored by vLLM -- the model determines the behavior
|
||||
parallel_tool_calls: Optional[bool] = False
|
||||
|
||||
Reference in New Issue
Block a user