Module Utils¶

opacus.utils.module_utils.are_state_dict_equal(sd1, sd2)[source]¶

Compares two state dicts, while logging discrepancies

opacus.utils.module_utils.clone_module(module)[source]¶

Handy utility to clone an nn.Module. PyTorch doesn’t always support copy.deepcopy(), so it is just easier to serialize the model to a BytesIO and read it from there. When weights_only=False, torch.load() uses “pickle” module implicity, which is known to be insecure. Only load the model you trust.

Parameters:

module (Module) – The module to clone

Return type:

Module

Returns:

The clone of module

opacus.utils.module_utils.get_submodule(module, target)[source]¶

Returns the submodule given by target if it exists, otherwise throws an error.

This is copy-pasta of Pytorch 1.9’s get_submodule() implementation; and is included here to also support Pytorch 1.8. This function can be removed in favour of module.get_submodule() once Opacus abandons support for torch 1.8.

See more details at https://pytorch.org/docs/stable/generated/torch.nn.Module.html?highlight=get_submodule#torch.nn.Module.get_submodule

Parameters:

• module (Module) – module

• target (str) – submodule string

Return type:

Module

Returns:

The submodule given by target if it exists

Raises:

AttributeError – If submodule doesn’t exist

opacus.utils.module_utils.parametrized_modules(module)[source]¶

Recursively iterates over all submodules, returning those that have parameters (as opposed to “wrapper modules” that just organize modules).

Return type:

Iterable[Tuple[str, Module]]

opacus.utils.module_utils.requires_grad(module, *, recurse=False)[source]¶

Checks if any parameters in a specified module require gradients.

Parameters:

• module (Module) – PyTorch module whose parameters are to be examined.

• recurse (bool) – Flag specifying if the gradient requirement check should be applied recursively to submodules of the specified module

Return type:

bool

Returns:

Flag indicate if any parameters require gradients

opacus.utils.module_utils.trainable_modules(module)[source]¶

Recursively iterates over all submodules, returning those that have parameters and are trainable (ie they want a grad).

Return type:

Iterable[Tuple[str, Module]]

opacus.utils.module_utils.trainable_parameters(module)[source]¶

Recursively iterates over all parameters, returning those that are trainable (ie they want a grad).

Return type:

Iterable[Tuple[str, Parameter]]