biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[GptOss] Add GptOss reasoning parser to support structure output (#22322)

Browse Source 
Signed-off-by: Chen Zhang <zhangch99@outlook.com>
Co-authored-by: LiuXiaoxuanPKU <lilyliupku@gmail.com>
Co-authored-by: simon-mo <xmo@berkeley.edu>
Co-authored-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
Co-authored-by: Hongxia Yang <62075498+hongxiayang@users.noreply.github.com>
Co-authored-by: Minseok Lee <47620120+minseokl@users.noreply.github.com>
Co-authored-by: Yongye Zhu <zyy1102000@gmail.com>
This commit is contained in:
Chen Zhang
2025-08-05 23:39:13 -07:00
committed by GitHub
parent 98a3a81024
commit a47e6ffe93
3 changed files with 69 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
vllm/model_executor/models/config.py
View File

@@ -247,13 +247,13 @@ class GraniteMoeHybridModelConfig(VerifyAndUpdateConfig):
config.max_model_len)
class GptOssConfig(VerifyAndUpdateConfig):
class GptOssForCausalLMConfig(VerifyAndUpdateConfig):
@staticmethod
def verify_and_update_config(vllm_config: "VllmConfig") -> None:
decoding_config = vllm_config.decoding_config
if decoding_config.reasoning_backend == "":
decoding_config.reasoning_backend = "openai"
decoding_config.reasoning_backend = "GptOss"
# Increase the max capture size from 512 to 1024 for performance.
# NOTE(woosuk): This will increase the number of CUDA graphs
@@ -373,5 +373,5 @@ MODELS_CONFIG_MAP: dict[str, type[VerifyAndUpdateConfig]] = {
"JinaVLForRanking": JinaVLForSequenceClassificationConfig,
"JambaForSequenceClassification": JambaForSequenceClassificationConfig,
"GraniteMoeHybridForCausalLM": GraniteMoeHybridModelConfig,
"GptOssForCausalLM": GptOssConfig,
"GptOssForCausalLM": GptOssForCausalLMConfig,
}