onnx_diagnostic.helpers.args_helper

onnx_diagnostic.helpers.args_helper.check_cuda_availability()

Checks if CUDA is available without pytorch or onnxruntime. Calls nvidia-smi.

onnx_diagnostic.helpers.args_helper.get_parsed_args(name: str, scenarios: Dict[str, str] | None = None, description: str | None = None, epilog: str | None = None, number: int = 10, repeat: int = 10, warmup: int = 5, sleep: float = 0.1, tries: int = 2, expose: str | None = None, new_args: List[str] | None = None, **kwargs: Dict[str, Tuple[int | str | float, str]]) → Namespace

Returns parsed arguments for examples in this package.

Parameters:

• name – script name

• scenarios – list of available scenarios

• description – parser description

• epilog – text at the end of the parser

• number – default value for number parameter

• repeat – default value for repeat parameter

• warmup – default value for warmup parameter

• sleep – default value for sleep parameter

• expose – if empty, keeps all the parameters, if not None, only publish kwargs contains, otherwise the list of parameters to publish separated by a comma

• new_args – args to consider or None to take sys.args

• kwargs – additional parameters, example: n_trees=(10, “number of trees to train”)

Returns:

parser