2023-04-26 06:37:25 +00:00
|
|
|
import warnings
|
2023-05-09 03:10:02 +00:00
|
|
|
from typing import Callable, Iterator, List, Optional, Tuple, Union
|
2023-04-26 06:37:25 +00:00
|
|
|
|
|
|
|
import torch
|
|
|
|
import torch.nn as nn
|
|
|
|
from torch import Tensor
|
|
|
|
from torch.optim import Optimizer
|
|
|
|
from torch.optim.lr_scheduler import _LRScheduler as LRScheduler
|
|
|
|
from torch.utils._pytree import tree_map
|
|
|
|
from torch.utils.data import DataLoader
|
|
|
|
|
2023-05-19 11:42:31 +00:00
|
|
|
from colossalai.checkpoint_io import CheckpointIO, GeneralCheckpointIO
|
2023-04-26 06:37:25 +00:00
|
|
|
from colossalai.interface import ModelWrapper, OptimizerWrapper
|
|
|
|
from colossalai.utils import get_current_device
|
|
|
|
from colossalai.zero import zero_model_wrapper, zero_optim_wrapper
|
|
|
|
|
2023-05-05 11:36:10 +00:00
|
|
|
from .dp_plugin_base import DPPluginBase
|
2023-04-26 06:37:25 +00:00
|
|
|
from .torch_ddp_plugin import TorchDDPCheckpointIO
|
|
|
|
|
|
|
|
__all__ = ['LowLevelZeroPlugin']
|
|
|
|
|
|
|
|
|
|
|
|
def _convert_to_fp16(x):
|
|
|
|
if isinstance(x, torch.Tensor) and torch.is_floating_point(x):
|
|
|
|
return x.half()
|
|
|
|
return x
|
|
|
|
|
|
|
|
|
|
|
|
class LowLevelZeroCheckpointIO(TorchDDPCheckpointIO):
|
|
|
|
|
|
|
|
def save_unsharded_optimizer(self, optimizer: Optimizer, checkpoint: str, gather_dtensor: bool):
|
|
|
|
"""
|
|
|
|
Save optimizer to checkpoint but only on master process.
|
|
|
|
"""
|
2023-05-19 11:42:31 +00:00
|
|
|
# TODO(ver217): optimizer state dict is sharded, and cannot get full state dict now
|
|
|
|
warnings.warn(
|
|
|
|
'LowLevelZeroPlugin does not support save full optimizer checkpoint now. Save it on every process.')
|
|
|
|
checkpoint = f'{checkpoint}.rank{self.coordinator.rank}'
|
|
|
|
GeneralCheckpointIO.save_unsharded_optimizer(self, optimizer, checkpoint, gather_dtensor)
|
|
|
|
|
|
|
|
def load_optimizer(self, optimizer: Optimizer, checkpoint: str):
|
|
|
|
warnings.warn(
|
|
|
|
'LowLevelZeroPlugin can only load optimizer checkpoint saved by itself with the same number of processes.')
|
|
|
|
checkpoint = f'{checkpoint}.rank{self.coordinator.rank}'
|
|
|
|
super().load_optimizer(optimizer, checkpoint)
|
2023-04-26 06:37:25 +00:00
|
|
|
|
|
|
|
|
|
|
|
class LowLevelZeroModel(ModelWrapper):
|
|
|
|
|
|
|
|
def __init__(self, module: nn.Module, stage: int, precision: str) -> None:
|
|
|
|
super().__init__(module)
|
|
|
|
self.convert_inputs = (precision == 'fp16')
|
|
|
|
module = zero_model_wrapper(module, zero_stage=stage)
|
|
|
|
if precision == 'fp16':
|
|
|
|
module = module.half()
|
|
|
|
module = module.to(get_current_device())
|
|
|
|
self.module = module
|
|
|
|
|
|
|
|
def forward(self, *args, **kwargs):
|
|
|
|
if self.convert_inputs:
|
|
|
|
args = tree_map(_convert_to_fp16, args)
|
|
|
|
kwargs = tree_map(_convert_to_fp16, kwargs)
|
|
|
|
return super().forward(*args, **kwargs)
|
|
|
|
|
|
|
|
|
|
|
|
class LowLevelZeroOptimizer(OptimizerWrapper):
|
|
|
|
|
|
|
|
def __init__(self,
|
|
|
|
module: nn.Module,
|
|
|
|
optimizer: Optimizer,
|
|
|
|
zero_optim_config: dict,
|
|
|
|
optim_kwargs: dict,
|
|
|
|
verbose: bool = False) -> None:
|
|
|
|
optimizer = zero_optim_wrapper(module,
|
|
|
|
optimizer,
|
|
|
|
optim_config=zero_optim_config,
|
|
|
|
**optim_kwargs,
|
|
|
|
verbose=verbose)
|
|
|
|
super().__init__(optimizer)
|
|
|
|
|
|
|
|
def backward(self, loss: Tensor, *args, **kwargs):
|
|
|
|
self.optim.backward(loss)
|
|
|
|
|
|
|
|
def clip_grad_by_norm(self,
|
|
|
|
max_norm: Union[float, int],
|
|
|
|
norm_type: Union[float, int] = 2,
|
|
|
|
error_if_nonfinite: bool = False,
|
|
|
|
*args,
|
|
|
|
**kwargs) -> Tensor:
|
|
|
|
warnings.warn(f'LowLevelZero controls grad clipping by itself, so you should not use clip_grad_by_norm')
|
|
|
|
|
|
|
|
def clip_grad_by_value(self, clip_value: float, *args, **kwargs) -> None:
|
|
|
|
raise NotImplementedError('LowLevelZero does not support clip_grad_by_value')
|
|
|
|
|
|
|
|
|
2023-05-05 11:36:10 +00:00
|
|
|
class LowLevelZeroPlugin(DPPluginBase):
|
2023-04-26 06:37:25 +00:00
|
|
|
"""
|
|
|
|
Plugin for low level zero.
|
|
|
|
|
|
|
|
Example:
|
|
|
|
>>> from colossalai.booster import Booster
|
|
|
|
>>> from colossalai.booster.plugin import LowLevelZeroPlugin
|
|
|
|
>>>
|
|
|
|
>>> model, train_dataset, optimizer, criterion = ...
|
|
|
|
>>> plugin = LowLevelZeroPlugin()
|
|
|
|
|
2023-05-08 07:44:03 +00:00
|
|
|
>>> train_dataloader = plugin.prepare_dataloader(train_dataset, batch_size=8)
|
2023-04-26 06:37:25 +00:00
|
|
|
>>> booster = Booster(plugin=plugin)
|
|
|
|
>>> model, optimizer, train_dataloader, criterion = booster.boost(model, optimizer, train_dataloader, criterion)
|
|
|
|
|
|
|
|
Args:
|
|
|
|
strage (int, optional): ZeRO stage. Defaults to 1.
|
|
|
|
precision (str, optional): precision. Support 'fp16' and 'fp32'. Defaults to 'fp16'.
|
|
|
|
initial_scale (float, optional): Initial scale used by DynamicGradScaler. Defaults to 2**32.
|
|
|
|
min_scale (float, optional): Min scale used by DynamicGradScaler. Defaults to 1.
|
|
|
|
growth_factor (float, optional): growth_factor used by DynamicGradScaler. Defaults to 2.
|
|
|
|
backoff_factor (float, optional): backoff_factor used by DynamicGradScaler. Defaults to 0.5.
|
|
|
|
growth_interval (float, optional): growth_interval used by DynamicGradScaler. Defaults to 1000.
|
|
|
|
hysteresis (float, optional): hysteresis used by DynamicGradScaler. Defaults to 2.
|
|
|
|
max_scale (int, optional): max_scale used by DynamicGradScaler. Defaults to 2**32.
|
|
|
|
max_norm (float, optional): max_norm used for `clip_grad_norm`. You should notice that you shall not do
|
|
|
|
clip_grad_norm by yourself when using ZeRO DDP. The ZeRO optimizer will take care of clip_grad_norm.
|
|
|
|
norm_type (float, optional): norm_type used for `clip_grad_norm`.
|
|
|
|
reduce_bucket_size_in_m (int, optional): grad reduce bucket size in M. Defaults to 12.
|
|
|
|
communication_dtype (torch.dtype, optional): communication dtype. If not specified, the dtype of param will be used. Defaults to None.
|
|
|
|
overlap_communication (bool, optional): whether to overlap communication and computation. Defaults to True.
|
|
|
|
cpu_offload (bool, optional): whether to offload grad, master weight and optimizer state to cpu. Defaults to False.
|
|
|
|
verbose (bool, optional): verbose mode. Debug info including grad overflow will be printed. Defaults to False.
|
|
|
|
"""
|
|
|
|
|
|
|
|
def __init__(
|
|
|
|
self,
|
|
|
|
stage: int = 1,
|
|
|
|
precision: str = 'fp16',
|
|
|
|
initial_scale: float = 2**32,
|
|
|
|
min_scale: float = 1,
|
|
|
|
growth_factor: float = 2,
|
|
|
|
backoff_factor: float = 0.5,
|
|
|
|
growth_interval: int = 1000,
|
|
|
|
hysteresis: int = 2,
|
|
|
|
max_scale: float = 2**32,
|
|
|
|
max_norm: float = 0.0,
|
|
|
|
norm_type: float = 2.0,
|
|
|
|
reduce_bucket_size_in_m: int = 12,
|
|
|
|
communication_dtype: Optional[torch.dtype] = None,
|
|
|
|
overlap_communication: bool = True,
|
|
|
|
cpu_offload: bool = False,
|
|
|
|
verbose: bool = False,
|
|
|
|
) -> None:
|
2023-05-05 11:36:10 +00:00
|
|
|
super().__init__()
|
2023-04-26 06:37:25 +00:00
|
|
|
assert stage in (1, 2), f'LowLevelZeroPlugin only supports stage 1/2 training'
|
|
|
|
assert precision in ('fp16', 'fp32'), f'LowLevelZeroPlugin only supports fp16/fp32 training'
|
|
|
|
|
|
|
|
self.stage = stage
|
|
|
|
self.precision = precision
|
|
|
|
self.zero_optim_config = dict(reduce_bucket_size=reduce_bucket_size_in_m * 1024 * 1024,
|
|
|
|
communication_dtype=communication_dtype,
|
|
|
|
overlap_communication=overlap_communication,
|
|
|
|
cpu_offload=cpu_offload)
|
|
|
|
self.optim_kwargs = dict(initial_scale=initial_scale,
|
|
|
|
growth_factor=growth_factor,
|
|
|
|
backoff_factor=backoff_factor,
|
|
|
|
growth_interval=growth_interval,
|
|
|
|
hysteresis=hysteresis,
|
|
|
|
min_scale=min_scale,
|
|
|
|
max_scale=max_scale,
|
|
|
|
max_norm=max_norm,
|
|
|
|
norm_type=norm_type)
|
|
|
|
self.verbose = verbose
|
|
|
|
|
|
|
|
def support_no_sync(self) -> bool:
|
|
|
|
return False
|
|
|
|
|
|
|
|
def control_precision(self) -> bool:
|
|
|
|
return True
|
|
|
|
|
|
|
|
def supported_precisions(self) -> List[str]:
|
|
|
|
return ['fp16', 'fp32']
|
|
|
|
|
|
|
|
def control_device(self) -> bool:
|
|
|
|
return True
|
|
|
|
|
|
|
|
def supported_devices(self) -> List[str]:
|
|
|
|
return ['cuda']
|
|
|
|
|
|
|
|
def configure(
|
|
|
|
self,
|
|
|
|
model: nn.Module,
|
|
|
|
optimizer: Optimizer,
|
|
|
|
criterion: Callable = None,
|
|
|
|
dataloader: DataLoader = None,
|
|
|
|
lr_scheduler: LRScheduler = None,
|
|
|
|
) -> Tuple[Union[nn.Module, OptimizerWrapper, LRScheduler, DataLoader]]:
|
|
|
|
|
|
|
|
if not isinstance(model, ModelWrapper):
|
|
|
|
model = LowLevelZeroModel(model, self.stage, self.precision)
|
|
|
|
|
|
|
|
if not isinstance(optimizer, OptimizerWrapper):
|
|
|
|
optimizer = LowLevelZeroOptimizer(model.unwrap(), optimizer, self.zero_optim_config, self.optim_kwargs,
|
|
|
|
self.verbose)
|
|
|
|
|
|
|
|
return model, optimizer, criterion, dataloader, lr_scheduler
|
|
|
|
|
|
|
|
def control_checkpoint_io(self) -> bool:
|
|
|
|
return True
|
|
|
|
|
|
|
|
def get_checkpoint_io(self) -> CheckpointIO:
|
|
|
|
return LowLevelZeroCheckpointIO()
|
2023-05-09 03:10:02 +00:00
|
|
|
|
|
|
|
def no_sync(self, model: nn.Module) -> Iterator[None]:
|
|
|
|
raise NotImplementedError
|