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, 01:00 UTC
[Misc] unify variable for LLM instance v2 (#21356)
Signed-off-by: Andy Xie <andy.xning@gmail.com>
Ning Xie committed
Jul 22, 2025 at 21:32 UTC
0df4d9b06b15fa39eeb2d440e7742da93afd5e6c
1 file changed
+4
-4
tests/models/language/generation/test_gemma.py
+4
-4
@@ -15,13 +15,13 @@ def test_dummy_loader(vllm_runner, monkeypatch, model: str) -> None:
15
load_format="dummy",
16
) as llm:
17
if model == "google/gemma-3-4b-it":
18
- normalizers = llm.model.collective_rpc(
18
+ normalizers = llm.llm.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
21
+ config = llm.llm.llm_engine.model_config.hf_config.text_config
22
else:
23
- normalizers = llm.model.collective_rpc(
23
+ normalizers = llm.llm.collective_rpc(
24
lambda self: self.model_runner.model.model.normalizer.cpu(
25
).item())
26
- config = llm.model.llm_engine.model_config.hf_config
26
+ config = llm.llm.llm_engine.model_config.hf_config
27
assert np.allclose(normalizers, config.hidden_size**0.5, rtol=2e-3)