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

[New Model]: Support GteNewModelForSequenceClassification (#23524)

Browse Source 
Signed-off-by: wang.yuqi <noooop@126.com>
This commit is contained in:
wang.yuqi
2025-08-28 15:36:42 +08:00
committed by GitHub
parent 186aced5ff
commit 11a7fafaa8
13 changed files with 157 additions and 76 deletions
Download Patch File Download Diff File Expand all files Collapse all files

19
tests/models/language/pooling/test_mxbai_rerank.py
View File

@@ -10,12 +10,20 @@ from tests.conftest import HfRunner
from ...utils import LASTPoolingRerankModelInfo, RerankModelInfo
from .mteb_utils import mteb_test_rerank_models
mxbai_rerank_hf_overrides = {
"architectures": ["Qwen2ForSequenceClassification"],
"classifier_from_token": ["0", "1"],
"method": "from_2_way_softmax",
}
RERANK_MODELS = [
LASTPoolingRerankModelInfo("mixedbread-ai/mxbai-rerank-base-v2",
architecture="Qwen2ForSequenceClassification",
hf_overrides=mxbai_rerank_hf_overrides,
enable_test=True),
LASTPoolingRerankModelInfo("mixedbread-ai/mxbai-rerank-large-v2",
architecture="Qwen2ForSequenceClassification",
hf_overrides=mxbai_rerank_hf_overrides,
enable_test=False)
]
@@ -71,13 +79,4 @@ class MxbaiRerankerHfRunner(HfRunner):
@pytest.mark.parametrize("model_info", RERANK_MODELS)
def test_rerank_models_mteb(vllm_runner, model_info: RerankModelInfo) -> None:
vllm_extra_kwargs: dict[str, Any] = {}
if model_info.architecture == "Qwen2ForSequenceClassification":
vllm_extra_kwargs["hf_overrides"] = {
"architectures": ["Qwen2ForSequenceClassification"],
"classifier_from_token": ["0", "1"],
"method": "from_2_way_softmax",
}
mteb_test_rerank_models(MxbaiRerankerHfRunner, vllm_runner, model_info,
vllm_extra_kwargs)
mteb_test_rerank_models(MxbaiRerankerHfRunner, vllm_runner, model_info)