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

[Frontend] Support using chat template as custom score template for reranking models (#30550)

Browse Source 
Signed-off-by: Jakub Zakrzewski <jzakrzewski@nvidia.com>
Signed-off-by: wang.yuqi <yuqi.wang@daocloud.io>
Signed-off-by: wang.yuqi <noooop@126.com>
Co-authored-by: wang.yuqi <yuqi.wang@daocloud.io>
This commit is contained in:
Jakub Zakrzewski
2025-12-23 12:19:16 +01:00
committed by GitHub
parent 27c6c2f98c
commit 23daef548d
19 changed files with 663 additions and 46 deletions
Download Patch File Download Diff File Expand all files Collapse all files

22
vllm/model_executor/models/config.py
View File

@@ -88,6 +88,26 @@ class JinaRobertaModelConfig(VerifyAndUpdateConfig):
}
class LlamaBidirectionalConfig(VerifyAndUpdateConfig):
@staticmethod
def verify_and_update_config(vllm_config: "VllmConfig") -> None:
from vllm.config.pooler import PoolingTypeStr
hf_config = vllm_config.model_config.hf_config
hf_config.is_causal = False
pooling_type_map: dict[str, PoolingTypeStr] = {
"avg": "MEAN",
"cls": "CLS",
"last": "LAST",
}
pooling_type = pooling_type_map.get(hf_config.pooling, None)
if pooling_type is None:
raise ValueError(f"pool_type {hf_config.pooling} not supported")
vllm_config.model_config.pooler_config.pooling_type = pooling_type
class NomicBertModelConfig(VerifyAndUpdateConfig):
@staticmethod
def verify_and_update_config(vllm_config: "VllmConfig") -> None:
@@ -509,6 +529,8 @@ MODELS_CONFIG_MAP: dict[str, type[VerifyAndUpdateConfig]] = {
"GteNewModel": GteNewModelConfig,
"GteNewForSequenceClassification": GteNewModelConfig,
"Gemma3TextModel": Gemma3TextModelConfig,
"LlamaBidirectionalForSequenceClassification": LlamaBidirectionalConfig,
"LlamaBidirectionalModel": LlamaBidirectionalConfig,
"NomicBertModel": NomicBertModelConfig,
"Qwen2ForProcessRewardModel": Qwen2ForProcessRewardModelConfig,
"Qwen2ForRewardModel": Qwen2ForRewardModelConfig,