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 15, 17:15 UTC
[Bugfix] Fix missing return value in load_weights method of adapters.py (#15542)
Signed-off-by: noc-turne <2270929247@qq.com>
Mrm committed
Mar 31, 2025 at 21:56 UTC
037bcd942cac24a30dfe61ea3423caf62c92a153
1 file changed
+5
-4
vllm/model_executor/models/adapters.py
+5
-4
@@ -99,16 +99,17 @@ def _create_pooling_model_cls(
99
mapper = WeightsMapper(orig_to_new_prefix={"model.": ""})
100
weights = mapper.apply(weights)
101
102
- self.model.load_weights(weights)
103
- return
102
+ loaded_params = self.model.load_weights(weights)
103
+ loaded_params = {f"model.{name}" for name in loaded_params}
104
+ return loaded_params
105
106
# For most other models
107
if hasattr(orig_cls, "load_weights"):
107
- orig_cls.load_weights(self, weights) # type: ignore
108
+ return orig_cls.load_weights(self, weights) # type: ignore
109
# Fallback
110
else:
111
loader = AutoWeightsLoader(self)
111
- loader.load_weights(weights)
112
+ return loader.load_weights(weights)
113
114
return ModelForPooling # type: ignore
115