A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 15, 04:45 UTC
[Fix] Fix gemma CI test failing on main (#20124)
Signed-off-by: Thomas Parnell <tpa@zurich.ibm.com>
Thomas Parnell committed
Jun 27, 2025 at 06:06 UTC
e11093068043a780ba8e778cdfcff8291d3f5b8c
1 file changed
+18
-11
tests/models/language/generation/test_gemma.py
+18
-11
@@ -7,14 +7,21 @@ MODELS = ["google/gemma-2b", "google/gemma-2-2b", "google/gemma-3-4b-it"]
7
8
9
@pytest.mark.parametrize("model", MODELS)
10
-def test_dummy_loader(vllm_runner, model: str) -> None:
11
- with vllm_runner(
12
- model,
13
- load_format="dummy",
14
- ) as llm:
15
- normalizers = llm.collective_rpc(lambda self: self.worker.model_runner.
16
- model.model.normalizer.cpu().item())
17
- assert np.allclose(
18
- normalizers,
19
- llm.llm_engine.model_config.hf_config.hidden_size**0.5,
20
- rtol=1e-3)
10
+def test_dummy_loader(vllm_runner, monkeypatch, model: str) -> None:
11
+ with monkeypatch.context() as m:
12
+ m.setenv("VLLM_ALLOW_INSECURE_SERIALIZATION", "1")
13
+ with vllm_runner(
14
+ model,
15
+ load_format="dummy",
16
+ ) as llm:
17
+ if model == "google/gemma-3-4b-it":
18
+ normalizers = llm.model.collective_rpc(
19
+ lambda self: self.model_runner.model.language_model.model.
20
+ normalizer.cpu().item())
21
+ config = llm.model.llm_engine.model_config.hf_config.text_config
22
+ else:
23
+ normalizers = llm.model.collective_rpc(
24
+ lambda self: self.model_runner.model.model.normalizer.cpu(
25
+ ).item())
26
+ config = llm.model.llm_engine.model_config.hf_config
27
+ assert np.allclose(normalizers, config.hidden_size**0.5, rtol=2e-3)