使用分布式检查点 (DCP) 的异步保存

Created On: Jul 22, 2024 | Last Updated: Jul 22, 2024 | Last Verified: Nov 05, 2024

作者： Lucas Pasqualin, Iris Zhang, Rodrigo Kumpera, Chien-Chin Huang

检查点通常是分布式训练任务关键路径中的瓶颈，随着模型和世界规模的增长，其成本也越来越高。一种优秀的策略是异步并行检查点。下面，我们扩展了 分布式检查点教程入门 中保存的示例，以展示如何非常轻松地与 torch.distributed.checkpoint.async_save 集成。

What you will learn

• 如何使用 DCP 并行生成检查点

• 优化性能的有效策略

Prerequisites

• PyTorch v2.4.0 或更高版本

• 分布式检查点教程入门

异步检查点概述

在开始使用异步检查点之前，了解它与同步检查点之间的差异和限制很重要。具体来说：

• 内存需求 - 异步检查点通过将模型首先复制到内部CPU缓冲区来工作。

  这非常有帮助，因为它确保模型和优化器权重在检查点过程中不会改变，但会将CPU内存提高至“checkpoint_size_per_rank X number_of_ranks”的倍数。此外，用户应注意了解其系统的内存限制。特别是，固定内存在使用“页面锁定”内存，这种内存相比于“可分页”内存更加稀缺。

• 检查点管理 - 由于检查点是异步的，用户需要自行管理同时运行的检查点。通常，用户可以

  通过处理 async_save 返回的未来对象来采用自己的管理策略。对于大多数用户，我们建议将检查点限制为一次一个异步请求，以避免因每次请求带来的额外内存压力。

import os

import torch
import torch.distributed as dist
import torch.distributed.checkpoint as dcp
import torch.multiprocessing as mp
import torch.nn as nn

from torch.distributed.fsdp import FullyShardedDataParallel as FSDP
from torch.distributed.checkpoint.state_dict import get_state_dict, set_state_dict
from torch.distributed.checkpoint.stateful import Stateful
from torch.distributed.fsdp.fully_sharded_data_parallel import StateDictType

CHECKPOINT_DIR = "checkpoint"


class AppState(Stateful):
    """This is a useful wrapper for checkpointing the Application State. Since this object is compliant
    with the Stateful protocol, DCP will automatically call state_dict/load_stat_dict as needed in the
    dcp.save/load APIs.

    Note: We take advantage of this wrapper to hande calling distributed state dict methods on the model
    and optimizer.
    """

    def __init__(self, model, optimizer=None):
        self.model = model
        self.optimizer = optimizer

    def state_dict(self):
        # this line automatically manages FSDP FQN's, as well as sets the default state dict type to FSDP.SHARDED_STATE_DICT
        model_state_dict, optimizer_state_dict = get_state_dict(model, optimizer)
        return {
            "model": model_state_dict,
            "optim": optimizer_state_dict
        }

    def load_state_dict(self, state_dict):
        # sets our state dicts on the model and optimizer, now that we've loaded
        set_state_dict(
            self.model,
            self.optimizer,
            model_state_dict=state_dict["model"],
            optim_state_dict=state_dict["optim"]
        )

class ToyModel(nn.Module):
    def __init__(self):
        super(ToyModel, self).__init__()
        self.net1 = nn.Linear(16, 16)
        self.relu = nn.ReLU()
        self.net2 = nn.Linear(16, 8)

    def forward(self, x):
        return self.net2(self.relu(self.net1(x)))


def setup(rank, world_size):
    os.environ["MASTER_ADDR"] = "localhost"
    os.environ["MASTER_PORT"] = "12355 "

    # initialize the process group
    dist.init_process_group("nccl", rank=rank, world_size=world_size)
    torch.cuda.set_device(rank)


def cleanup():
    dist.destroy_process_group()


def run_fsdp_checkpoint_save_example(rank, world_size):
    print(f"Running basic FSDP checkpoint saving example on rank {rank}.")
    setup(rank, world_size)

    # create a model and move it to GPU with id rank
    model = ToyModel().to(rank)
    model = FSDP(model)

    loss_fn = nn.MSELoss()
    optimizer = torch.optim.Adam(model.parameters(), lr=0.1)

    checkpoint_future = None
    for step in range(10):
        optimizer.zero_grad()
        model(torch.rand(8, 16, device="cuda")).sum().backward()
        optimizer.step()

        # waits for checkpointing to finish if one exists, avoiding queuing more then one checkpoint request at a time
        if checkpoint_future is not None:
            checkpoint_future.result()

        state_dict = { "app": AppState(model, optimizer) }
        checkpoint_future = dcp.async_save(state_dict, checkpoint_id=f"{CHECKPOINT_DIR}_step{step}")

    cleanup()


if __name__ == "__main__":
    world_size = torch.cuda.device_count()
    print(f"Running async checkpoint example on {world_size} devices.")
    mp.spawn(
        run_fsdp_checkpoint_save_example,
        args=(world_size,),
        nprocs=world_size,
        join=True,
    )

使用固定内存获得更高性能

如果上述优化仍然不足以满足性能需求，您可以通过使用额外优化来利用固定内存缓冲区对GPU模型进行检查点处理。特别地，这种优化针对异步检查点的主要开销，即将内存复制到检查点缓冲区。通过在检查点请求之间保持一个固定内存缓冲区，用户可以利用直接内存访问来加速复制。

备注

这种优化的主要缺点是在检查点步骤之间保持缓冲区的持久性。没有固定内存优化（如上所述），任何检查点缓冲区在检查点完成后都会被释放。使用固定内存实现时，此缓冲区在步骤之间保持存在，导致应用生命周期内持续承受同样的峰值内存压力。

import os

import torch
import torch.distributed as dist
import torch.distributed.checkpoint as dcp
import torch.multiprocessing as mp
import torch.nn as nn

from torch.distributed.fsdp import FullyShardedDataParallel as FSDP
from torch.distributed.checkpoint.state_dict import get_state_dict, set_state_dict
from torch.distributed.checkpoint.stateful import Stateful
from torch.distributed.fsdp.fully_sharded_data_parallel import StateDictType
from torch.distributed.checkpoint import StorageWriter

CHECKPOINT_DIR = "checkpoint"


class AppState(Stateful):
    """This is a useful wrapper for checkpointing the Application State. Since this object is compliant
    with the Stateful protocol, DCP will automatically call state_dict/load_stat_dict as needed in the
    dcp.save/load APIs.

    Note: We take advantage of this wrapper to hande calling distributed state dict methods on the model
    and optimizer.
    """

    def __init__(self, model, optimizer=None):
        self.model = model
        self.optimizer = optimizer

    def state_dict(self):
        # this line automatically manages FSDP FQN's, as well as sets the default state dict type to FSDP.SHARDED_STATE_DICT
        model_state_dict, optimizer_state_dict = get_state_dict(model, optimizer)
        return {
            "model": model_state_dict,
            "optim": optimizer_state_dict
        }

    def load_state_dict(self, state_dict):
        # sets our state dicts on the model and optimizer, now that we've loaded
        set_state_dict(
            self.model,
            self.optimizer,
            model_state_dict=state_dict["model"],
            optim_state_dict=state_dict["optim"]
        )

class ToyModel(nn.Module):
    def __init__(self):
        super(ToyModel, self).__init__()
        self.net1 = nn.Linear(16, 16)
        self.relu = nn.ReLU()
        self.net2 = nn.Linear(16, 8)

    def forward(self, x):
        return self.net2(self.relu(self.net1(x)))


def setup(rank, world_size):
    os.environ["MASTER_ADDR"] = "localhost"
    os.environ["MASTER_PORT"] = "12355 "

    # initialize the process group
    dist.init_process_group("nccl", rank=rank, world_size=world_size)
    torch.cuda.set_device(rank)


def cleanup():
    dist.destroy_process_group()


def run_fsdp_checkpoint_save_example(rank, world_size):
    print(f"Running basic FSDP checkpoint saving example on rank {rank}.")
    setup(rank, world_size)

    # create a model and move it to GPU with id rank
    model = ToyModel().to(rank)
    model = FSDP(model)

    loss_fn = nn.MSELoss()
    optimizer = torch.optim.Adam(model.parameters(), lr=0.1)

    # The storage writer defines our 'staging' strategy, where staging is considered the process of copying
    # checkpoints to in-memory buffers. By setting `cached_state_dict=True`, we enable efficient memory copying
    # into a persistent buffer with pinned memory enabled.
    # Note: It's important that the writer persists in between checkpointing requests, since it maintains the
    # pinned memory buffer.
    writer = StorageWriter(cached_state_dict=True)
    checkpoint_future = None
    for step in range(10):
        optimizer.zero_grad()
        model(torch.rand(8, 16, device="cuda")).sum().backward()
        optimizer.step()

        state_dict = { "app": AppState(model, optimizer) }
        if checkpoint_future is not None:
            # waits for checkpointing to finish, avoiding queuing more then one checkpoint request at a time
            checkpoint_future.result()
        dcp.async_save(state_dict, storage_writer=writer, checkpoint_id=f"{CHECKPOINT_DIR}_step{step}")

    cleanup()


if __name__ == "__main__":
    world_size = torch.cuda.device_count()
    print(f"Running fsdp checkpoint example on {world_size} devices.")
    mp.spawn(
        run_fsdp_checkpoint_save_example,
        args=(world_size,),
        nprocs=world_size,
        join=True,
    )

结论

总之，我们学习了如何使用DCP的 async_save API 在关键训练路径之外生成检查点。我们还了解了使用此API引入的额外内存和并发开销，以及利用固定内存进一步加速的额外优化。

• 保存和加载模型教程

• FullyShardedDataParallel教程介绍