load_model_from_path() — pytorch Function Reference

Architecture documentation for the load_model_from_path() function in common.py from the pytorch codebase.

Function python DynamoBenchmarks Microbenchmarking calls 1 called by 1

Entity Profile

DynamoBenchmarks→ Microbenchmarking→ load_model_from_path() — pytorch Function Reference

Dependency Diagram

graph TD
  6233fde7_853e_8fb3_0ae6_703b798d8c5d["load_model_from_path()"]
  c9be2096_e6d7_2374_ad2e_a6e33f435ada["run()"]
  c9be2096_e6d7_2374_ad2e_a6e33f435ada -->|calls| 6233fde7_853e_8fb3_0ae6_703b798d8c5d
  ce8fd365_4112_b289_9c73_7345d5e35203["RuntimeError()"]
  6233fde7_853e_8fb3_0ae6_703b798d8c5d -->|calls| ce8fd365_4112_b289_9c73_7345d5e35203
  style 6233fde7_853e_8fb3_0ae6_703b798d8c5d fill:#6366f1,stroke:#818cf8,color:#fff

Relationship Graph

Source Code

benchmarks/dynamo/common.py lines 285–320

def load_model_from_path(path_and_class_str):
    configs = {}
    for kvstr in path_and_class_str.split(","):
        k, v = kvstr.split(":")
        configs[k] = v

    for name in ["path", "class"]:
        if name not in configs:
            raise RuntimeError(
                "Invalid --only arguments. Check help message for the correct format"
            )

    path = configs["path"]
    class_name = configs["class"]

    if path[:1] != "/":
        raise RuntimeError(
            "Use absolute path since dynamo may change the current working directory which makes using relative path tricky"
        )

    spec = importlib.util.spec_from_file_location("module_name", path)
    module = importlib.util.module_from_spec(spec)
    spec.loader.exec_module(module)

    model_class = getattr(module, class_name)
    if not issubclass(model_class, torch.nn.Module):
        raise AssertionError(
            f"expected {class_name} to be a subclass of torch.nn.Module, got {model_class}"
        )
    model = model_class()
    if not hasattr(model, "get_example_inputs"):
        raise AssertionError(
            f"expected model {class_name} to have get_example_inputs method"
        )
    inputs = model.get_example_inputs()
    return model, inputs