SingleDeviceStrategy

class lightning.pytorch.strategies.SingleDeviceStrategy(device='cpu', accelerator=None, checkpoint_io=None, precision_plugin=None)[source]

Bases: Strategy

Strategy that handles communication on a single device.

all_gather(tensor, group=None, sync_grads=False)[source]

Perform a all_gather on all processes.

Return type:

Tensor

barrier(*args, **kwargs)[source]

Synchronizes all processes which blocks processes until the whole group enters this function.

Parameters:

name – an optional name to pass into barrier.

Return type:

None

broadcast(obj, src=0)[source]

Broadcasts an object to all processes.

Parameters:
  • obj (TypeVar(TBroadcast)) – the object to broadcast

  • src (int) – source rank

Return type:

TypeVar(TBroadcast)

model_to_device()[source]

Moves the model to the correct device.

Return type:

None

reduce(tensor, *args, **kwargs)[source]

Reduces a tensor from several distributed processes to one aggregated tensor. Since this strategy only operates with a single device, the reduction is simply the identity.

Parameters:
  • tensor (Any | Tensor) – the tensor to sync and reduce

  • *args (Any) – ignored

  • **kwargs (Any) – ignored

Return type:

Any | Tensor

Returns:

the unmodified input as reduction is not needed for single process operation

property is_global_zero: bool

Whether the current process is the rank zero process not only on the local node, but for all nodes.

property root_device: device

Returns the root device.