Source code for pytorch_lightning.plugins.training_type.horovod
# Copyright The PyTorch Lightning team.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from contextlib import ExitStack
from typing import Any, List, Optional, Tuple, Union
import torch
import torch.nn as nn
from torch.optim import Optimizer
from torch.optim.lr_scheduler import _LRScheduler
from pytorch_lightning.core.optimizer import LightningOptimizer
from pytorch_lightning.plugins.training_type.parallel import ParallelPlugin
from pytorch_lightning.utilities import _HOROVOD_AVAILABLE
from pytorch_lightning.utilities.distributed import distributed_available
from pytorch_lightning.utilities.distributed import group as dist_group
from pytorch_lightning.utilities.distributed import rank_zero_only, ReduceOp
if _HOROVOD_AVAILABLE:
import horovod.torch as hvd
[docs]class HorovodPlugin(ParallelPlugin):
"""Plugin for Horovod distributed training integration."""
def __init__(self, parallel_devices: Optional[List[torch.device]] = None):
super().__init__(parallel_devices=parallel_devices, cluster_environment=None)
rank_zero_only.rank = self.global_rank
@property
def global_rank(self) -> int:
return hvd.rank()
@property
def local_rank(self) -> int:
return hvd.local_rank()
@property
def world_size(self) -> int:
return hvd.size()
@property
def root_device(self):
return self.parallel_devices[self.local_rank]
@property
def distributed_sampler_kwargs(self):
distributed_sampler_kwargs = dict(num_replicas=self.world_size, rank=self.global_rank)
return distributed_sampler_kwargs
[docs] def pre_dispatch(self):
if not self.lightning_module.trainer.training:
# no need to setup optimizers
return
def _unpack_lightning_optimizer(opt):
return opt._optimizer if isinstance(opt, LightningOptimizer) else opt
optimizers = self.lightning_module.trainer.optimizers
optimizers = [_unpack_lightning_optimizer(opt) for opt in optimizers]
# Horovod: scale the learning rate by the number of workers to account for
# increased total batch size
for optimizer in optimizers:
for param_group in optimizer.param_groups:
param_group["lr"] *= self.world_size
# Horovod: adjust base LR used by schedulers to match scaled optimizer initial LR
lr_schedulers = self.lightning_module.trainer.lr_schedulers
for scheduler in lr_schedulers:
scheduler = scheduler["scheduler"]
if isinstance(scheduler, _LRScheduler):
scheduler.base_lrs = [lr * self.world_size for lr in scheduler.base_lrs]
# Horovod: broadcast parameters & optimizer state to ensure consistent initialization
hvd.broadcast_parameters(self.lightning_module.state_dict(), root_rank=0)
for optimizer in optimizers:
hvd.broadcast_optimizer_state(optimizer, root_rank=0)
self.lightning_module.trainer.accelerator.optimizers = self._wrap_optimizers(optimizers)
def start_training(self, trainer):
with ExitStack() as stack:
for optimizer in trainer.optimizers:
# Synchronization will be performed explicitly following backward()
stack.enter_context(optimizer.skip_synchronize())
# set up training routine
self._results = trainer.run_stage()
# Make sure all workers have finished training before returning to the user
self.join()
def start_evaluating(self, trainer):
with ExitStack():
self._results = trainer.run_stage()
# Make sure all workers have finished training before returning to the user
self.join()
def start_predicting(self, trainer):
with ExitStack():
# set up training routine
self._results = trainer.run_stage()
# Make sure all workers have finished training before returning to the user
self.join()
[docs] def broadcast(self, obj: object, src: int = 0) -> object:
obj = hvd.broadcast_object(obj, src)
return obj
[docs] def model_to_device(self):
if self.on_gpu:
# this can potentially be removed after #8312. Not done due to lack of horovod testing
torch.cuda.set_device(self.root_device)
self.model.to(self.root_device)
def join(self):
if self.on_gpu:
hvd.join(self.local_rank)
else:
hvd.join()
[docs] def reduce(self, tensor, group: Optional[Any] = None, reduce_op: Optional[Union[ReduceOp, str]] = "mean"):
"""
Reduces a tensor from several distributed processes to one aggregated tensor.
Args:
tensor: the tensor to sync and reduce
group: the process group to gather results from. Defaults to all processes (world)
reduce_op: the reduction operation. Defaults to 'mean'/'avg'.
Can also be a string 'sum' to calculate the sum during reduction.
Return:
reduced value, except when the input was not a tensor the output remains is unchanged
"""
if group is not None:
raise ValueError("Horovod does not support allreduce using a subcommunicator at this time. Unset `group`.")
if reduce_op in (None, "avg", "mean"):
reduce_op = hvd.Average
elif reduce_op in ("sum", ReduceOp.SUM):
reduce_op = hvd.Sum
else:
raise ValueError(f"unrecognized `reduce_op`: {reduce_op}")
# sync all processes before reduction
self.join()
return hvd.allreduce(tensor, op=reduce_op)
[docs] def all_gather(
self, result: Union[torch.Tensor], group: Optional[Any] = dist_group.WORLD, sync_grads: bool = False
) -> torch.Tensor:
if group is not None and group != dist_group.WORLD:
raise ValueError("Horovod does not support allgather using a subcommunicator at this time. Unset `group`.")
if len(result.shape) == 0:
# Convert scalars to single dimension tensors
result = result.reshape(1)
# sync and gather all
self.join()
gathered = hvd.allgather(result)
gathered_result = list(gathered.split(1, dim=0))
return gathered_result
[docs] def post_backward(self, closure_loss: torch.Tensor) -> None:
# synchronize all horovod optimizers.
for optimizer in self.lightning_module.trainer.optimizers:
optimizer.synchronize()
def _wrap_optimizers(self, optimizers: List[Optimizer]) -> List["hvd.DistributedOptimizer"]:
"""Wraps optimizers to perform gradient aggregation via allreduce."""
return [
hvd.DistributedOptimizer(opt, named_parameters=self._filter_named_parameters(self.lightning_module, opt))
if "horovod" not in str(opt.__class__)
else opt
for opt in optimizers
]
@staticmethod
def _filter_named_parameters(model: nn.Module, optimizer: Optimizer) -> List[Tuple[str, nn.Parameter]]:
opt_params = {p for group in optimizer.param_groups for p in group.get("params", [])}
return [(name, p) for name, p in model.named_parameters() if p in opt_params]
