From 94f68426fa186b03a201f8bfb3ff34b1b1387f6c Mon Sep 17 00:00:00 2001 From: Thomas Parnell Date: Thu, 26 Jun 2025 11:51:00 +0000 Subject: [PATCH 1/2] Fix gemma test Signed-off-by: Thomas Parnell --- .../models/language/generation/test_gemma.py | 25 +++++++++++-------- 1 file changed, 14 insertions(+), 11 deletions(-) diff --git a/tests/models/language/generation/test_gemma.py b/tests/models/language/generation/test_gemma.py index ed0f0c19a041..3d4ceb17a623 100644 --- a/tests/models/language/generation/test_gemma.py +++ b/tests/models/language/generation/test_gemma.py @@ -7,14 +7,17 @@ @pytest.mark.parametrize("model", MODELS) -def test_dummy_loader(vllm_runner, model: str) -> None: - with vllm_runner( - model, - load_format="dummy", - ) as llm: - normalizers = llm.collective_rpc(lambda self: self.worker.model_runner. - model.model.normalizer.cpu().item()) - assert np.allclose( - normalizers, - llm.llm_engine.model_config.hf_config.hidden_size**0.5, - rtol=1e-3) +def test_dummy_loader(vllm_runner, monkeypatch, model: str) -> None: + with monkeypatch.context() as m: + m.setenv("VLLM_ALLOW_INSECURE_SERIALIZATION", "1") + with vllm_runner( + model, + load_format="dummy", + ) as llm: + normalizers = llm.model.collective_rpc( + lambda self: self.model_runner.model.model.normalizer.cpu( + ).item()) + assert np.allclose( + normalizers, + llm.model.llm_engine.model_config.hf_config.hidden_size**0.5, + rtol=1e-3) From 7efba289cf07806bbc2080f06e54e7ae124c9154 Mon Sep 17 00:00:00 2001 From: Thomas Parnell Date: Thu, 26 Jun 2025 15:19:35 +0000 Subject: [PATCH 2/2] Handle gemma-3-4b-it correctly Signed-off-by: Thomas Parnell --- tests/models/language/generation/test_gemma.py | 18 +++++++++++------- 1 file changed, 11 insertions(+), 7 deletions(-) diff --git a/tests/models/language/generation/test_gemma.py b/tests/models/language/generation/test_gemma.py index 3d4ceb17a623..5be4ae874e61 100644 --- a/tests/models/language/generation/test_gemma.py +++ b/tests/models/language/generation/test_gemma.py @@ -14,10 +14,14 @@ def test_dummy_loader(vllm_runner, monkeypatch, model: str) -> None: model, load_format="dummy", ) as llm: - normalizers = llm.model.collective_rpc( - lambda self: self.model_runner.model.model.normalizer.cpu( - ).item()) - assert np.allclose( - normalizers, - llm.model.llm_engine.model_config.hf_config.hidden_size**0.5, - rtol=1e-3) + if model == "google/gemma-3-4b-it": + normalizers = llm.model.collective_rpc( + lambda self: self.model_runner.model.language_model.model. + normalizer.cpu().item()) + config = llm.model.llm_engine.model_config.hf_config.text_config + else: + normalizers = llm.model.collective_rpc( + lambda self: self.model_runner.model.model.normalizer.cpu( + ).item()) + config = llm.model.llm_engine.model_config.hf_config + assert np.allclose(normalizers, config.hidden_size**0.5, rtol=2e-3)