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

[Refactor] Separate sequence and token pooling types (#32026)

Browse Source 
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
This commit is contained in:
Cyrus Leung
2026-01-10 12:53:24 +08:00
committed by GitHub
parent 52d428295d
commit 583a90e005
42 changed files with 324 additions and 204 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
vllm/model_executor/models/roberta.py
View File

@@ -93,7 +93,7 @@ class RobertaClassificationHead(nn.Module):
return x
@default_pooling_type("CLS")
@default_pooling_type(seq_pooling_type="CLS")
class RobertaEmbeddingModel(BertEmbeddingModel):
"""A model that uses Roberta to provide embedding functionalities."""
@@ -150,7 +150,7 @@ class RobertaEmbeddingModel(BertEmbeddingModel):
return loader.load_weights(weights_list, mapper=mapper)
@default_pooling_type("CLS")
@default_pooling_type(seq_pooling_type="CLS")
class RobertaForSequenceClassification(nn.Module, SupportsCrossEncoding):
"""A model that uses Roberta to provide embedding functionalities.