returnn.torch.util.module

Utils for modules

returnn.torch.util.module.convert_parameters_to_buffers(module: Module, parameter_names: Collection[str], *, deep: bool = True, persistent: bool)

Parameters:

• module

• parameter_names

• deep – parameter_name can contain ‘.’ to access submodules

• persistent – whether the buffer is persistent. if True, the buffer will be saved to the state_dict. passed to module.register_buffer.

returnn.torch.util.module.convert_parameter_to_buffer(module: Module, parameter_name: str, *, deep: bool = True, persistent: bool)

Parameters:

• module

• parameter_name

• deep – parameter_name can contain ‘.’ to access submodules

• persistent – whether the buffer is persistent. if True, the buffer will be saved to the state_dict. passed to module.register_buffer.