[ModelRunner v2] Support Transformers pooling model (#52425)

Signed-off-by: Taneem Ibrahim <taneem.ibrahim@gmail.com>
This commit is contained in:
Taneem Ibrahim
2026-08-16 15:25:49 -04:00
committed by GitHub
parent ef43e3101b
commit fdab2b10bc
2 changed files with 18 additions and 5 deletions
+10 -1
View File
@@ -263,7 +263,16 @@ def test_embed_loading(vllm_runner, model):
@pytest.mark.parametrize(
"arch", ["TransformersEmbeddingModel", "TransformersForSequenceClassification"]
)
def test_pooling(hf_runner, vllm_runner, example_prompts, arch):
@pytest.mark.parametrize("use_v2_model_runner", [False, True], ids=["v1", "v2"])
def test_pooling(
hf_runner,
vllm_runner,
example_prompts,
arch,
monkeypatch,
use_v2_model_runner,
):
monkeypatch.setenv("VLLM_USE_V2_MODEL_RUNNER", str(int(use_v2_model_runner)))
model = get_model(arch)
vllm_kwargs = dict(max_model_len=None, model_impl="transformers")