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

[Model][3/N] Automatic conversion of CrossEncoding model (#20168)

Browse Source 
Signed-off-by: wang.yuqi <noooop@126.com>
This commit is contained in:
wang.yuqi
2025-07-04 20:47:39 +08:00
committed by GitHub
parent 9e5452ee34
commit 2e26f9156a
8 changed files with 234 additions and 133 deletions
Download Patch File Download Diff File Expand all files Collapse all files

16
tests/models/language/pooling/test_gte.py
View File

@@ -56,10 +56,16 @@ MODELS = [
enable_test=False),
]
V1FlashAttentionImpNotSupported = [
"Alibaba-NLP/gte-Qwen2-1.5B-instruct", "Alibaba-NLP/gte-modernbert-base"
]
@pytest.mark.parametrize("model_info", MODELS)
def test_embed_models_mteb(hf_runner, vllm_runner,
model_info: EmbedModelInfo) -> None:
def test_embed_models_mteb(hf_runner, vllm_runner, model_info: EmbedModelInfo,
monkeypatch) -> None:
if model_info.name in V1FlashAttentionImpNotSupported:
monkeypatch.setenv("VLLM_USE_V1", "0")
vllm_extra_kwargs: dict[str, Any] = {}
if model_info.architecture == "GteNewModel":
@@ -71,8 +77,10 @@ def test_embed_models_mteb(hf_runner, vllm_runner,
@pytest.mark.parametrize("model_info", MODELS)
def test_embed_models_correctness(hf_runner, vllm_runner,
model_info: EmbedModelInfo,
example_prompts) -> None:
model_info: EmbedModelInfo, example_prompts,
monkeypatch) -> None:
if model_info.name in V1FlashAttentionImpNotSupported:
monkeypatch.setenv("VLLM_USE_V1", "0")
vllm_extra_kwargs: dict[str, Any] = {}
if model_info.architecture == "GteNewModel":