[Fronten] Clean up stop_token_ids override for Harmony (#44009)

Signed-off-by: Yifan Zong <yzong@redhat.com>
This commit is contained in:
yzong-rh
2026-05-29 13:28:06 -07:00
committed by GitHub
parent 38b864d81d
commit 46409fd2a1
4 changed files with 0 additions and 23 deletions
@@ -57,7 +57,6 @@ from vllm.entrypoints.openai.engine.serving import (
)
from vllm.entrypoints.openai.models.serving import OpenAIServingModels
from vllm.entrypoints.openai.parser.harmony_utils import (
get_stop_tokens_for_assistant_actions,
get_streamable_parser_for_assistant,
parse_chat_output,
)
@@ -158,13 +157,6 @@ class OpenAIServingChat(OpenAIServing):
else getattr(mc, "override_generation_config", {}).get("max_new_tokens")
)
self.use_harmony = self.model_config.hf_config.model_type == "gpt_oss"
if self.use_harmony:
if "stop_token_ids" not in self.default_sampling_params:
self.default_sampling_params["stop_token_ids"] = []
self.default_sampling_params["stop_token_ids"].extend(
get_stop_tokens_for_assistant_actions()
)
self.tool_call_id_type = get_tool_call_id_type(self.model_config)
# NOTE(woosuk): While OpenAI's chat completion API supports browsing