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

[Model][2/N] Improve all pooling task | Support multi-vector retrieval (#25370)

Browse Source 
Signed-off-by: wang.yuqi <noooop@126.com>
This commit is contained in:
wang.yuqi
2025-10-15 19:14:41 +08:00
committed by GitHub
parent d4d1a6024f
commit f54f85129e
41 changed files with 786 additions and 399 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
vllm/model_executor/models/qwen2_rm.py
View File

@@ -107,7 +107,7 @@ class Qwen2ForRewardModel(Qwen2RewardBaseModel):
assert pooler_config is not None
self.pooler = DispatchPooler(
{"encode": Pooler.for_encode(pooler_config)},
{"token_classify": Pooler.for_token_classify(pooler_config)}
)
@@ -120,4 +120,6 @@ class Qwen2ForProcessRewardModel(Qwen2RewardBaseModel):
pooler_config = vllm_config.model_config.pooler_config
assert pooler_config is not None
self.pooler = DispatchPooler({"encode": Pooler.for_encode(pooler_config)})
self.pooler = DispatchPooler(
{"token_classify": Pooler.for_token_classify(pooler_config)}
)