vathos.utils

The various utility functions that are essential for vathos

load_config(filename: str) → dict

Load a configuration file as YAML and returns a dict

Parameters:filename (str) – location of the file Returns:(Dict) of the config

setup_device(model: torch.nn.modules.module.Module, target_device: int) → Tuple[torch.device, List[int]]

sets up the device for the model

Parameters:

• model (nn.Module) – the model
• target_device (int) – index of the target device

Returns:

Tuple[torch.device, List[int]]

setup_param_groups(model: torch.nn.modules.module.Module, config: Dict) → List

get_instance(module: module, name: str, config: Dict, *args) → Any

creates an instance from a constructor name and module name

Parameters:

• module (ModuleType) – the module which contains the class
• name (str) – name of the class
• config (Dict) – configuration of experiment
• args (Any) – any arguments that needs to be passed to the class

get_instance_v2(module, ctor_name, *args, **kwargs)

creates an instance from a constructor name and module name

Parameters:

• module – the module which contains the ctor_name
• ctor_name – name of the constructor
• args (Optional) – positional arguments that needs to be passed to ctor
• kwargs (Optional) – keywords arguments that needs to be passed to ctor

setup_logger(name)