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

[Model] Pooling models default to using chunked prefill & prefix caching if supported. (#20930)

Browse Source 
Signed-off-by: wang.yuqi <noooop@126.com>
This commit is contained in:
wang.yuqi
2025-08-12 00:41:37 +08:00
committed by GitHub
parent 16fb668b61
commit 84cf78acee
31 changed files with 452 additions and 261 deletions
Download Patch File Download Diff File Expand all files Collapse all files

15
tests/models/language/pooling/test_mxbai_rerank.py
View File

@@ -7,15 +7,16 @@ import torch
from tests.conftest import HfRunner
from .mteb_utils import RerankModelInfo, mteb_test_rerank_models
from ...utils import LASTPoolingRerankModelInfo, RerankModelInfo
from .mteb_utils import mteb_test_rerank_models
RERANK_MODELS = [
RerankModelInfo("mixedbread-ai/mxbai-rerank-base-v2",
architecture="Qwen2ForSequenceClassification",
enable_test=True),
RerankModelInfo("mixedbread-ai/mxbai-rerank-large-v2",
architecture="Qwen2ForSequenceClassification",
enable_test=False)
LASTPoolingRerankModelInfo("mixedbread-ai/mxbai-rerank-base-v2",
architecture="Qwen2ForSequenceClassification",
enable_test=True),
LASTPoolingRerankModelInfo("mixedbread-ai/mxbai-rerank-large-v2",
architecture="Qwen2ForSequenceClassification",
enable_test=False)
]