A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 10, 04:00 UTC
Support configure skip_special_tokens in openai response api (#32345)
Signed-off-by: Shiyan Deng <dsy842974287@meta.com>
Shiyan Deng committed
Jan 14, 2026 at 20:07 UTC
375e5984fec8f79f1ec4190c2fd76cc185f6a58f
1 file changed
+2
vllm/entrypoints/openai/responses/protocol.py
+2
@@ -171,6 +171,7 @@ class ResponsesRequest(OpenAIBaseModel):
171
top_k: int | None = None
172
truncation: Literal["auto", "disabled"] | None = "disabled"
173
user: str | None = None
174
+ skip_special_tokens: bool = True
175
176
# --8<-- [start:responses-extra-params]
177
request_id: str = Field(
@@ -279,6 +280,7 @@ class ResponsesRequest(OpenAIBaseModel):
280
structured_outputs=structured_outputs,
281
logit_bias=self.logit_bias,
282
skip_clone=True, # Created fresh per request, safe to skip clone
283
+ skip_special_tokens=self.skip_special_tokens,
284
)
285
286
def is_include_output_logprobs(self) -> bool: