MCPcopy
hub / github.com/NVIDIA/TensorRT-LLM / llama_7b_bs2_path

Function llama_7b_bs2_path

tests/unittest/llmapi/test_executor.py:53–65  ·  view source on GitHub ↗
(engine_path: Path)

Source from the content-addressed store, hash-verified

51
52@pytest.fixture(scope="module")
53def llama_7b_bs2_path(engine_path: Path) -> Path:
54 path = engine_path / "llama7b_bs2"
55
56 if not path.exists():
57 model_dir = str(llm_models_root() / "llama-models/llama-7b-hf")
58 build_config = BuildConfig()
59 build_config.max_beam_width = 2
60 # TODO[chunweiy]: switch to executor backend
61 llm = LLM(model_dir, build_config=build_config)
62 with llm:
63 llm.save(str(path))
64
65 return path
66
67
68@pytest.fixture(scope="module")

Callers

nothing calls this directly

Calls 4

llm_models_rootFunction · 0.90
BuildConfigClass · 0.50
LLMClass · 0.50
saveMethod · 0.45

Tested by

no test coverage detected