pytorch_pfn_extras.nn.parallel.DistributedDataParallel

class pytorch_pfn_extras.nn.parallel.DistributedDataParallel(module, broadcast_buffers=True, negotiate_grads=True, process_group=None, reduce_function=None, broadcast_function=None, **kwargs)

Bases: Module

Module for distributed data parallelism

This class synchronizes the gradients and the buffers after backward computations.

Parameters:

• module (Module) – torch.nn.Module object to be trained

• broadcast_buffers (bool) – Boolean flag to broadcast buffers after backward computations. Broadcasting buffers may be helpful when the module includes BatchNormalization. However, it will degrade training throughput. (default: True)

• negotiate_grads (bool) – Boolean flag to choose gradients to be sent before all-reduce. This flag is necessary when the computation graph of the module is dynamic. (default: True)

• process_group (Optional[ProcessGroup]) – Process group used for broadcasting and reducing. (default: torch.distributed.group.WORLD)

• reduce_function (Optional[Callable[[Sequence[Tensor], Optional[ProcessGroup]], None]]) – All-reduce function

• broadcast_function (Optional[Callable[[Sequence[Tensor], Optional[ProcessGroup]], None]]) – Broadcast function

• kwargs (Any) –

This module receives keyword arguments for the compatibility with torch.nn.parallel.DistributedDataParallel. It shows a warning when setting the ignored arguments.

Methods

__init__(module[, broadcast_buffers, ...])	This module receives keyword arguments for the compatibility with torch.nn.parallel.DistributedDataParallel.
add_module(name, module)	Add a child module to the current module.
apply(fn)	Apply fn recursively to every submodule (as returned by .children()) as well as self.
bfloat16()	Casts all floating point parameters and buffers to bfloat16 datatype.
buffers([recurse])	Return an iterator over module buffers.
children()	Return an iterator over immediate children modules.
compile(*args, **kwargs)	Compile this Module's forward using torch.compile().
cpu()	Move all model parameters and buffers to the CPU.
cuda([device])	Move all model parameters and buffers to the GPU.
double()	Casts all floating point parameters and buffers to double datatype.
eval()	Set the module in evaluation mode.
extra_repr()	Set the extra representation of the module.
float()	Casts all floating point parameters and buffers to float datatype.
forward(*args, **kwargs)	Define the computation performed at every call.
get_buffer(target)	Return the buffer given by target if it exists, otherwise throw an error.
get_extra_state()	Return any extra state to include in the module's state_dict.
get_parameter(target)	Return the parameter given by target if it exists, otherwise throw an error.
get_submodule(target)	Return the submodule given by target if it exists, otherwise throw an error.
half()	Casts all floating point parameters and buffers to half datatype.
ipu([device])	Move all model parameters and buffers to the IPU.
load_state_dict(state_dict[, strict])	Copy parameters and buffers from state_dict into this module and its descendants.
modules()	Return an iterator over all modules in the network.
named_buffers([prefix, recurse, ...])	Return an iterator over module buffers, yielding both the name of the buffer as well as the buffer itself.
named_children()	Return an iterator over immediate children modules, yielding both the name of the module as well as the module itself.
named_modules([memo, prefix, remove_duplicate])	Return an iterator over all modules in the network, yielding both the name of the module as well as the module itself.
named_parameters([prefix, recurse, ...])	Return an iterator over module parameters, yielding both the name of the parameter as well as the parameter itself.
no_sync()	A context manager to disable synchronization after backward
parameters([recurse])	Return an iterator over module parameters.
register_backward_hook(hook)	Register a backward hook on the module.
register_buffer(name, tensor[, persistent])	Add a buffer to the module.
register_comm_hook(hook)	Registers a hook function.
register_forward_hook(hook, *[, prepend, ...])	Register a forward hook on the module.
register_forward_pre_hook(hook, *[, ...])	Register a forward pre-hook on the module.
register_full_backward_hook(hook[, prepend])	Register a backward hook on the module.
register_full_backward_pre_hook(hook[, prepend])	Register a backward pre-hook on the module.
register_load_state_dict_post_hook(hook)	Register a post hook to be run after module's load_state_dict is called.
register_module(name, module)	Alias for add_module().
register_parameter(name, param)	Add a parameter to the module.
register_state_dict_pre_hook(hook)	Register a pre-hook for the load_state_dict() method.
requires_grad_([requires_grad])	Change if autograd should record operations on parameters in this module.
set_extra_state(state)	Set extra state contained in the loaded state_dict.
share_memory()	See torch.Tensor.share_memory_().
state_dict()	Return a dictionary containing references to the whole state of the module.
to(*args, **kwargs)	Move and/or cast the parameters and buffers.
to_empty(*, device[, recurse])	Move the parameters and buffers to the specified device without copying storage.
train([mode])	Set the module in training mode.
type(dst_type)	Casts all parameters and buffers to dst_type.
xpu([device])	Move all model parameters and buffers to the XPU.
zero_grad([set_to_none])	Reset gradients of all model parameters.

Attributes

T_destination	alias of TypeVar('T_destination', bound=Mapping[str, Tensor])
call_super_init
dump_patches

T_destination

alias of TypeVar(‘T_destination’, bound=Mapping[str, Tensor])

__init__(module, broadcast_buffers=True, negotiate_grads=True, process_group=None, reduce_function=None, broadcast_function=None, **kwargs)

This module receives keyword arguments for the compatibility with torch.nn.parallel.DistributedDataParallel. It shows a warning when setting the ignored arguments.

Parameters:

• module (Module) –

• broadcast_buffers (bool) –

• negotiate_grads (bool) –

• process_group (Optional[ProcessGroup]) –

• reduce_function (Optional[Callable[[Sequence[Tensor], Optional[ProcessGroup]], None]]) –

• broadcast_function (Optional[Callable[[Sequence[Tensor], Optional[ProcessGroup]], None]]) –

• kwargs (Any) –

Return type:

None

forward(*args, **kwargs)

Define the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

Parameters:

• args (Any) –

• kwargs (Any) –

Return type:

Any

load_state_dict(state_dict, strict=True, *args)

Copy parameters and buffers from state_dict into this module and its descendants.

If strict is True, then the keys of state_dict must exactly match the keys returned by this module’s state_dict() function.

Warning

If assign is True the optimizer must be created after the call to load_state_dict.

Parameters:

• state_dict (dict) – a dict containing parameters and persistent buffers.

• strict (bool, optional) – whether to strictly enforce that the keys in state_dict match the keys returned by this module’s state_dict() function. Default: True

• assign (bool, optional) – whether to assign items in the state dictionary to their corresponding keys in the module instead of copying them inplace into the module’s current parameters and buffers. When False, the properties of the tensors in the current module are preserved while when True, the properties of the Tensors in the state dict are preserved. Default: False

• args (Any) –

Returns:

• missing_keys is a list of str containing the missing keys

• unexpected_keys is a list of str containing the unexpected keys

Return type:

NamedTuple with missing_keys and unexpected_keys fields

Note

If a parameter or buffer is registered as None and its corresponding key exists in state_dict, load_state_dict() will raise a RuntimeError.

no_sync()

A context manager to disable synchronization after backward

Return type:

Generator[None, None, None]

register_comm_hook(hook)

Registers a hook function. This module will invoke the hook before starting the synchronization.

Args: hook: Callable object that will be invoked before synchronization

Parameters:

hook (Callable[[DistributedDataParallel], None]) –

Return type:

RemovableHandle

state_dict()

Return a dictionary containing references to the whole state of the module.

Both parameters and persistent buffers (e.g. running averages) are included. Keys are corresponding parameter and buffer names. Parameters and buffers set to None are not included.

Note

The returned object is a shallow copy. It contains references to the module’s parameters and buffers.

Warning

Currently state_dict() also accepts positional arguments for destination, prefix and keep_vars in order. However, this is being deprecated and keyword arguments will be enforced in future releases.

Warning

Please avoid the use of argument destination as it is not designed for end-users.

Parameters:

• destination (dict, optional) – If provided, the state of module will be updated into the dict and the same object is returned. Otherwise, an OrderedDict will be created and returned. Default: None.

• prefix (str, optional) – a prefix added to parameter and buffer names to compose the keys in state_dict. Default: ''.

• keep_vars (bool, optional) – by default the Tensor s returned in the state dict are detached from autograd. If it’s set to True, detaching will not be performed. Default: False.

Returns:

a dictionary containing a whole state of the module

Return type:

dict

Example:

>>> # xdoctest: +SKIP("undefined vars")
>>> module.state_dict().keys()
['bias', 'weight']

training: bool