[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:
Isotr0py
2025-10-13 18:21:48 +08:00
committed by GitHub
parent 4073c82c4e
commit 8e67b2557a
2 changed files with 13 additions and 7 deletions

View File

@@ -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",
)
]