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

44
tests/models/language/pooling/test_intfloat.py
View File

@@ -2,34 +2,34 @@
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
import pytest
from ...utils import EmbedModelInfo
from ...utils import CLSPoolingEmbedModelInfo, EmbedModelInfo
from .embed_utils import correctness_test_embed_models
from .mteb_utils import mteb_test_embed_models
MODELS = [
########## BertModel
EmbedModelInfo("intfloat/e5-small",
architecture="BertModel",
enable_test=True),
EmbedModelInfo("intfloat/e5-base",
architecture="BertModel",
enable_test=False),
EmbedModelInfo("intfloat/e5-large",
architecture="BertModel",
enable_test=False),
EmbedModelInfo("intfloat/multilingual-e5-small",
architecture="BertModel",
enable_test=False),
CLSPoolingEmbedModelInfo("intfloat/e5-small",
architecture="BertModel",
enable_test=True),
CLSPoolingEmbedModelInfo("intfloat/e5-base",
architecture="BertModel",
enable_test=False),
CLSPoolingEmbedModelInfo("intfloat/e5-large",
architecture="BertModel",
enable_test=False),
CLSPoolingEmbedModelInfo("intfloat/multilingual-e5-small",
architecture="BertModel",
enable_test=False),
########## XLMRobertaModel
EmbedModelInfo("intfloat/multilingual-e5-base",
architecture="XLMRobertaModel",
enable_test=True),
EmbedModelInfo("intfloat/multilingual-e5-large",
architecture="XLMRobertaModel",
enable_test=False),
EmbedModelInfo("intfloat/multilingual-e5-large-instruct",
architecture="XLMRobertaModel",
enable_test=False),
CLSPoolingEmbedModelInfo("intfloat/multilingual-e5-base",
architecture="XLMRobertaModel",
enable_test=True),
CLSPoolingEmbedModelInfo("intfloat/multilingual-e5-large",
architecture="XLMRobertaModel",
enable_test=False),
CLSPoolingEmbedModelInfo("intfloat/multilingual-e5-large-instruct",
architecture="XLMRobertaModel",
enable_test=False),
]