HorovodStrategy¶

class pytorch_lightning.strategies.HorovodStrategy(accelerator=None, parallel_devices=None, checkpoint_io=None, precision_plugin=None)[source]¶

Bases: pytorch_lightning.strategies.parallel.ParallelStrategy

Plugin for Horovod distributed training integration.

all_gather(result, group=torch.distributed.group.WORLD, 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.

broadcast(obj, src=0)[source]¶

Broadcasts an object to all processes.

Parameters

obj¶ (object) – the object to broadcast
src¶ (int) – source rank

Return type

object

model_to_device()[source]¶: Moves the model to the correct device.

post_backward(closure_loss)[source]¶

Run after precision plugin executes backward.

Return type: None

reduce(tensor, group=None, reduce_op='mean')[source]¶

Reduces a tensor from several distributed processes to one aggregated tensor.

Parameters

tensor¶ – the tensor to sync and reduce
group¶ (Optional[Any]) – the process group to gather results from. Defaults to all processes (world)
reduce_op¶ (Union[ReduceOp, str, None]) – the reduction operation. Defaults to ‘mean’/’avg’. Can also be a string ‘sum’ to calculate the sum during reduction.

Returns

reduced value, except when the input was not a tensor the output remains is unchanged

setup(trainer)[source]¶

Setup plugins for the trainer fit and creates optimizers.

Parameters: trainer¶ (Trainer) – the trainer instance
Return type: None

teardown()[source]¶

This method is called to teardown the training process.

It is the right place to release memory and free other resources.

Return type: None

property handles_gradient_accumulation: bool¶

Whether the plugin handles gradient accumulation internally.

Return type: bool

property root_device¶: Return the root device.