File size: 11,150 Bytes
9c6594c |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 |
# Copyright The Lightning AI 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.
import itertools
from collections.abc import Iterable, Iterator, Sized
from typing import Any, Callable, Optional, Union, cast
import torch
from torch import Tensor
from torch.nn.parallel.distributed import DistributedDataParallel
from torch.utils.data import DistributedSampler, Sampler
from typing_extensions import Self, override
from lightning_fabric.utilities.distributed import _DatasetSamplerWrapper
from pytorch_lightning.utilities.rank_zero import rank_zero_debug, rank_zero_info
from pytorch_lightning.utilities.types import _SizedIterable
def _find_tensors(
obj: Union[Tensor, list, tuple, dict, Any],
) -> Union[list[Tensor], itertools.chain]: # pragma: no-cover
"""Recursively find all tensors contained in the specified object."""
if isinstance(obj, Tensor):
return [obj]
if isinstance(obj, (list, tuple)):
return itertools.chain(*map(_find_tensors, obj))
if isinstance(obj, dict):
return itertools.chain(*map(_find_tensors, obj.values()))
return []
# In manual_optimization, we need to call reducer prepare_for_backward.
# Note: Keep track of PyTorch DDP and update if there is a change
# https://github.com/pytorch/pytorch/blob/v2.0.0/torch/nn/parallel/distributed.py#L1163-L1178
def prepare_for_backward(model: DistributedDataParallel, output: Any) -> None:
# `prepare_for_backward` is `DistributedDataParallel` specific.
if torch.is_grad_enabled() and model.require_backward_grad_sync:
model.require_forward_param_sync = True
# We'll return the output object verbatim since it is a freeform
# object. We need to find any tensors in this object, though,
# because we need to figure out which parameters were used during
# this forward pass, to ensure we short circuit reduction for any
# unused parameters. Only if `find_unused_parameters` is set.
args = list(_find_tensors(output)) if model.find_unused_parameters and not model.static_graph else []
reducer = cast(torch._C._distributed_c10d.Reducer, model.reducer)
reducer._rebuild_buckets() # avoids "INTERNAL ASSERT FAILED" with `find_unused_parameters=False`
reducer.prepare_for_backward(args)
else:
model.require_forward_param_sync = False
def _register_ddp_comm_hook(
model: DistributedDataParallel,
ddp_comm_state: Optional[object] = None,
ddp_comm_hook: Optional[Callable] = None,
ddp_comm_wrapper: Optional[Callable] = None,
) -> None:
"""Function to register communication hook for DDP model https://pytorch.org/docs/master/ddp_comm_hooks.html.
Args:
model:
DDP model
ddp_comm_state:
state is passed to the hook and can be used to maintain
and update any state information that users would like to
maintain as part of the training process. Examples: error
feedback in gradient compression, peers to communicate with
next in GossipGrad etc.
ddp_comm_hook:
hook(state: object, bucket: dist._GradBucket) -> torch.futures.Future
This callable function is called once the bucket is ready. The
hook can perform whatever processing is needed and return
a Future indicating completion of any async work (ex: allreduce).
If the hook doesn't perform any communication, it can also
just return a completed Future. The Future should hold the
new value of grad bucket's tensors. Once a bucket is ready,
c10d reducer would call this hook and use the tensors returned
by the Future and copy grads to individual parameters.
ddp_comm_wrapper:
communication hook wrapper to support a communication hook such
as FP16 compression as wrapper, which could be combined with
ddp_comm_hook
Examples::
from torch.distributed.algorithms.ddp_comm_hooks import (
default_hooks as default,
powerSGD_hook as powerSGD,
post_localSGD_hook as post_localSGD,
)
# fp16_compress_hook for compress gradients
ddp_model = ...
_register_ddp_comm_hook(
model=ddp_model,
ddp_comm_hook=default.fp16_compress_hook,
)
# powerSGD_hook
ddp_model = ...
_register_ddp_comm_hook(
model=ddp_model,
ddp_comm_state=powerSGD.PowerSGDState(
process_group=None,
matrix_approximation_rank=1,
start_powerSGD_iter=5000,
),
ddp_comm_hook=powerSGD.powerSGD_hook,
)
# post_localSGD_hook
subgroup, _ = torch.distributed.new_subgroups()
ddp_model = ...
_register_ddp_comm_hook(
model=ddp_model,
state=post_localSGD.PostLocalSGDState(
process_group=None,
subgroup=subgroup,
start_localSGD_iter=1_000,
),
ddp_comm_hook=post_localSGD.post_localSGD_hook,
)
# fp16_compress_wrapper combined with other communication hook
ddp_model = ...
_register_ddp_comm_hook(
model=ddp_model,
ddp_comm_state=powerSGD.PowerSGDState(
process_group=None,
matrix_approximation_rank=1,
start_powerSGD_iter=5000,
),
ddp_comm_hook=powerSGD.powerSGD_hook,
ddp_comm_wrapper=default.fp16_compress_wrapper,
)
"""
if ddp_comm_hook is None:
return
# inform mypy that ddp_comm_hook is callable
ddp_comm_hook: Callable = ddp_comm_hook
if ddp_comm_wrapper is not None:
rank_zero_info(
f"DDP comm wrapper is provided, apply {ddp_comm_wrapper.__qualname__}({ddp_comm_hook.__qualname__})."
)
ddp_comm_hook = ddp_comm_wrapper(ddp_comm_hook)
rank_zero_debug(f"Registering DDP comm hook: {ddp_comm_hook.__qualname__}.")
model.register_comm_hook(state=ddp_comm_state, hook=ddp_comm_hook)
def _sync_module_states(module: torch.nn.Module) -> None:
"""Taken from https://github.com/pytorch/pytorch/blob/v2.0.0/torch/nn/parallel/distributed.py#L675-L682."""
parameters_to_ignore = set(getattr(module, "_ddp_params_and_buffers_to_ignore", []))
from torch.distributed.distributed_c10d import _get_default_group
from torch.distributed.utils import _sync_module_states as torch_sync_module_states
torch_sync_module_states(
module,
_get_default_group(),
250 * 1024 * 1024,
src=0,
params_and_buffers_to_ignore=parameters_to_ignore,
)
class UnrepeatedDistributedSampler(DistributedSampler):
"""A fork of the PyTorch DistributedSampler that doesn't repeat data, instead allowing the number of batches per
process to be off-by-one from each other. This makes this sampler usable for predictions (it's deterministic and
doesn't require shuffling). It is potentially unsafe to use this sampler for training, because during training the
DistributedDataParallel syncs buffers on each forward pass, so it could freeze if one of the processes runs one
fewer batch. During prediction, buffers are only synced on the first batch, so this is safe to use as long as each
process runs at least one batch. We verify this in an assert.
Taken from https://github.com/jpuigcerver/PyLaia/blob/v1.0.0/laia/data/unpadded_distributed_sampler.py and
https://github.com/pytorch/pytorch/issues/25162#issuecomment-634146002
"""
def __init__(self, *args: Any, **kwargs: Any) -> None:
super().__init__(*args, **kwargs)
if not isinstance(self.dataset, Sized):
raise TypeError("The given dataset must implement the `__len__` method.")
self.num_samples = len(range(self.rank, len(self.dataset), self.num_replicas))
self.total_size = len(self.dataset)
# If any process has at least one batch, every other process needs to
# have at least one batch, or the DistributedDataParallel could lock up.
assert self.num_samples >= 1 or self.total_size == 0
@override
def __iter__(self) -> Iterator[list[int]]:
if not isinstance(self.dataset, Sized):
raise TypeError("The given dataset must implement the `__len__` method.")
if self.shuffle:
# deterministically shuffle based on epoch
g = torch.Generator()
g.manual_seed(self.epoch)
indices = torch.randperm(len(self.dataset), generator=g).tolist()
else:
indices = list(range(len(self.dataset)))
assert len(indices) == self.total_size
# subsample
indices = indices[self.rank : self.total_size : self.num_replicas]
assert len(indices) == self.num_samples
return iter(indices)
class UnrepeatedDistributedSamplerWrapper(UnrepeatedDistributedSampler):
"""Equivalent class to ``DistributedSamplerWrapper`` but for the ``UnrepeatedDistributedSampler``."""
def __init__(self, sampler: Union[Sampler, Iterable], *args: Any, **kwargs: Any) -> None:
super().__init__(_DatasetSamplerWrapper(sampler), *args, **kwargs)
@override
def __iter__(self) -> Iterator:
self.dataset.reset()
return (self.dataset[index] for index in super().__iter__())
class _IndexBatchSamplerWrapper:
"""This class is used to wrap a :class:`torch.utils.data.BatchSampler` and capture its indices."""
def __init__(self, batch_sampler: _SizedIterable) -> None:
# do not call super().__init__() on purpose
self.seen_batch_indices: list[list[int]] = []
self.__dict__ = {
k: v
for k, v in batch_sampler.__dict__.items()
if k not in ("__next__", "__iter__", "__len__", "__getstate__")
}
self._batch_sampler = batch_sampler
self._iterator: Optional[Iterator[list[int]]] = None
def __next__(self) -> list[int]:
assert self._iterator is not None
batch = next(self._iterator)
self.seen_batch_indices.append(batch)
return batch
def __iter__(self) -> Self:
self.seen_batch_indices = []
self._iterator = iter(self._batch_sampler)
return self
def __len__(self) -> int:
return len(self._batch_sampler)
def __getstate__(self) -> dict[str, Any]:
state = self.__dict__.copy()
state["_iterator"] = None # cannot pickle 'generator' object
return state
|