[Bugfix] Fix out of bound index issue for Jina-embedding-v3 RoPE with cuda graph (#26687)
Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
This commit is contained in:
@@ -25,10 +25,6 @@ EMBEDDING_MODELS = [
|
||||
mteb_score=0.824413164,
|
||||
architecture="XLMRobertaModel",
|
||||
is_matryoshka=True,
|
||||
# The default max length of the model is 8194, which will crash
|
||||
# CUDAGraph due to odd length for Gemm. We set it to 8192 to avoid
|
||||
# avoid this issue.
|
||||
max_model_len=8192,
|
||||
dtype="float32",
|
||||
)
|
||||
]
|
||||
|
||||
Reference in New Issue
Block a user