Pytorch在1.9.0引入了torchrun，用其替代1.9.0以前版本的torch.distributed.launch。torchrun在torch.distributed.launch 功能的基础上主要新增了两个功能：
- Failover: 当worker训练失败时，会自动重新启动所有worker继续进行训练；
- Elastic: 可以动态增加或或删除node节点；
弹性训练代码同DDP代码编写的思路基本一致，只要在DDP代码上增加以下两点即可：
- checkpoint处理：由于再每次增加或删除node时，会将所有worker kill掉，然后再重新启动所有worker进行训练。因此，在训练代码中要对训练的状态进行保存，以保证重启后能接着上次的状态继续训练。
- 超参调解：由于node节点数的变化，会导致global batch size的变化，因此我们的learning rate一般也要做相应的调整，保证训练出的模质量不受影响。
代码见第二节 最下面
当编写完弹性训练代码后，我们可以使用torchrun来启动弹性训练任务：
- --nnodes=1:3 ：表示当前训练任务接受最少1个node，最多3个node参与分布式训练；
- --nproc_per_node=4:表示每个node上节点有4个process
- --max_restarts=3: worker group最大的重启次数；这里需要注意的是，node fail、node scale down和node scale up都会导致restart；
- --rdzv_id=1：一个unique的job id，所有node均使用同一个job id；
- --rdzv_backend: rendezvous的backend实现，默认支持c10d和etcd两种；rendezvous用于多个node之间的通信和协调；
- --rdzv_endpoint：rendezvous的地址，应该为一个node的host ip和port；
torchrun \
--nnodes=1:3\
--nproc_per_node=4\
--max_restarts=3\
--rdzv_id=1\
--rdzv_backend=c10d\
--rdzv_endpoint="192.0.0.1:1234"\
train_elastic.py3 整体架构
弹性调度的架构如上图所示，其中最关键角色为elastic agent。在每个Node上面都有一个elastic agent进程，其负责管理当前Node上面的所有workers。
当我们调用torchrun 命令启动弹性训练任务后：
- 首先，elastic agent会触发rendezvous 流程; rendezvous的功能是在所有elastic agent间做协调和同步，该接口会一直阻塞直到至少min个elastic agent加入进来后返回；
- 然后，elastic agent会启动当前Node的所有workers
- 最后，elastic agent会监控当前Node上所有workers的运行状态，并根据workers的状态进行相应的处理（例如restart worker）
4 Elastic Agent
本小结，我们详细分析下Elastic Agent的实现。Elastic Agent在Pytorch代码中由以下对象构成:
- Elastic Agent是抽象基类
- SimpleElasticAgent提供了更完整的Agent接口，并且实现了部分接口
- LocalElasticAgent则是实现剩余的接口
Elastic Agent在代码中的调用逻辑如下：
torch.distributed.launcher.api:launch_agent()弹性训练逻辑的入口；
- 首先、会构建一个RendezvousParameters来描述Rendezvous调用时所需要的参数，例如min_nodes/max_nodes/endpoint等；
- 然后、构建WorkerSpec描述当前Node上启动Wokers的信息， 例如max_restart/entrypoint等；
- 再然后，构建LocalElasticAgent对象；
- 最后，调用LocalElasticAgent的run接口启动当前node的workers进行弹性训练；
- Elastic run接口主要由两个部分逻辑组成：
- 若process group的状态为succeeded：调用_exit_barrier接口等待所有node上agent相应并退出
- 若process group的状态为unhealthy或failed: 如果重试次数小于_remaining_restart则restart所有worker进程，否则stop所有worker，并退出；
- 若process group的状态为healthy: 则判断当前是否有node等待加入，如果有则restart_worker；（注：restart worker的实现逻辑是先stop 所有worker，然后在调用_initialize_workers)
SimpleElasticAgent._initialize_workers：先调用_rendezvous等待至少min 个node加入，然后调用_start_workers接口在当前node上启动worker processwhile loop monitor worker：while循环，监控上一步启动process的状态
5 Rendezvous
5.1 基本概念
Pytorch中Rendezvous的实现涉及到很多概念，我们这里先把这些概念一一介绍下，然后再介绍Rendezvous的实现这样会清晰很多。
首先是_RendezvousState，每个ElasticAgent上都会存储一份_RendezvousState，并会在必要时进行彼此间的同步，_RendezvousState存储的内容如下:
- round: The current round of the rendezvous.
- complete: A boolean value indicating whether the current round of the rendezvous is complete.
- deadline: The time at which the current round of the rendezvous will be considered complete if it is still waiting for nodes to join.
- closed: A boolean value indicating whether the rendezvous is closed.
- participants: A dictionary of the participants and their corresponding ranks.
- wait_list:A set of nodes that are waiting to participate in the next round of the rendezvous.
- last_heartbeats: A dictionary containing each node's last heartbeat time.
那_RendezvousState是如何在所有ElasticAgent间进行同步的呢，Pytorch中又提出了Store的概念，在Pytorch中有TCPStore、FileStore和HashStore三种类型，在弹性训练场景，默认使用TCPStore。
TCPStore的典型用法如下：
- 其是一个典型的server-client架构，我们在process1上启动server，在proess2上启动client，通过TCPStore的set和get接口可以进行数据的设置和获取
- 在Rendezvous实现中即是通过TCPStore来对_RendezvousState进行设置和获取的。
import torch.distributed as dist
from datetime import timedelta
# Run on process 1 (server)
server_store = dist.TCPStore("127.0.0.1", 1234, 2, True, timedelta(seconds=30))
# Run on process 2 (client)
client_store = dist.TCPStore("127.0.0.1", 1234, 2, False)
# Use any of the store methods from either the client or server after initialization
server_store.set("first_key", "first_value")
client_store.get("first_key")Pytorch的Rendezvous实现中，通过C10dRendezvousBackend对TCPStore进行了封装，并提供了set_state和get_state接口，方便state的操作。(注：Pytorch中还提供了EtcdRendezvousBackend，该类型的RendezvousBackend通过Etcd来进行_RendezvousState的同步）。
C10dRendezvousBackend的主要实现如下，可以很清晰的看到get_state和set_state的实现，均是对store接口的调用.
class C10dRendezvousBackend(RendezvousBackend):
def get_state(self) -> Optional[Tuple[bytes, Token]]:
"""See base class."""
base64_state: bytes = self._call_store("get", self._key)
return self._decode_state(base64_state)
def set_state(
self, state: bytes, token: Optional[Token] = None
) -> Optional[Tuple[bytes, Token, bool]]:
"""See base class."""
base64_state_str: str = b64encode(state).decode()
if token:
# Shortcut if we know for sure that the token is not valid.
if not isinstance(token, bytes):
result = self.get_state()
if result is not None:
tmp = *result, False
# Python 3.6 does not support tuple unpacking in return
# statements.
return tmp
return None
token = token.decode()
else:
token = self._NULL_SENTINEL
base64_state: bytes = self._call_store("compare_set", self._key, token, base64_state_str)
state_token_pair = self._decode_state(base64_state)
if state_token_pair is None:
return None
new_state, new_token = state_token_pair
# C10d Store's compare_set method does not offer an easy way to find out
# whether our write attempt was successful. As a brute-force solution we
# perform a bitwise comparison of our local state and the remote state.
return new_state, new_token, new_state == state
def _call_store(self, store_op: str, *args, **kwargs) -> Any:
try:
return getattr(self._store, store_op)(*args, **kwargs)
except (ValueError, RuntimeError, TimeoutError) as exc:
raise RendezvousConnectionError(
"The connection to the C10d store has failed. See inner exception for details."
) from exc在RendezvousBackend的基础上，Pytorch提出了一个更偏向业务层面的概念**_RendezvousStateHolder**，其提供了_RendezvousState进行获取、同步、标记更新的接口，这些接口的实现均是调用RendezvousBackend的set_state和get_state完成的。
_RendezvousStateHolder的定义如下：
class _RendezvousStateHolder(ABC):
"""Holds the shared rendezvous state synced with other nodes."""
def state(self) -> _RendezvousState:
"""Gets the local state."""
def sync(self) -> Optional[bool]:
"""Reads or writes the latest state.
Returns:
A boolean value indicating whether the local state, in case marked
as dirty, was successfully synced with other nodes.
"""
def mark_dirty(self) -> None:
"""Marks the local state as dirty."""Rendezvous的基础设置都准备好了，状态在 _RendezvousState中保存，状态的同步通过 _RendezvousStateHolder来完成，此时还差一项，就是Rendezvous state的是如何变更的。这个变更通过 _RendezvousXXXOp和 _RendezvousOpExecutor共同来完成。
Pytorch首先提供了_RendezvousExitOp/_RendezvousJoinOp/_RendezvousCloseOp/_RendezvousKeepAliveOp来对应ElasticAgent的退出、加入、Rendezvous关闭和心跳保保持四个操作。这些OP的实现逻辑是根据OP的类型和当前_RendezvousState的内容来决定来返回一个action，_RendezvousOpExecutor则执行对应的action。
例如_RendezvousExitOp 对应ElasticAgent的退出操作
- 如果当前节点仍旧在participants列表中，则返回一个REMOVE_FROM_PARTICIPANTS，_RendezvousOpExecutor在接收到这个action后会执行_remove_from_participants逻辑；
- 如果当前节点没有在participants列表中，返回FINISH，这个状态_RendezvousOpExecutor不会做任何操作；
class _RendezvousExitOp:
"""Represents a rendezvous exit operation."""
def __call__(self, ctx: _RendezvousContext, deadline: float) -> _Action:
if ctx.node in ctx.state.participants:
if time.monotonic() > deadline:
return _Action.ERROR_TIMEOUT
return _Action.REMOVE_FROM_PARTICIPANTS
return _Action.FINISH_DistributedRendezvousOpExecutor的核心接口如下：
- run提供了执行Rendezvous op的总入口
- 其他接口则对应了Rendezvous op返回的action的实现。这些action的实现本质上都是对_RendezvousState内容的修改，例如_mark_rendezvous_closed是将_RendezvousState的close字段设置为了True。
class _DistributedRendezvousOpExecutor:
def run(self, state_handler: Callable[[_RendezvousContext, float], _Action], deadline: float,) -> None:
def _keep_alive(self) -> None:
def _add_to_participants(self)
def _add_to_wait_list(self)
def _remove_from_participants(self)
def _remove_from_wait_list(self)
def _mark_rendezvous_complete(self)
def _mark_rendezvous_closed(self):
self._state.closed = True最后一个要介绍的概念是RendezvousHandler，其是Rendezvous系统最上层的对外接口，ElasticAgent通过该接口来在所有节点间进行协调。在Pytorch中提供了DynamicRendezvousHandler、EtcdRendezvousHandler和StaticTCPRendezvous三种实现，这里我们仅关注DynamicRendezvousHandler。
RendezvousHandler中最核心的接口是next_rendezvous，ElasticAgent会调用该接口来等待至少min个node的加入。他们实现我们后面再进行讲解。
上面介绍的这些概念，可以通过如下的关系图来进行描述。
5.2 实现逻辑
在熟系完Rendezvous的基本概念后，我们现在可以来看其实现逻辑了。
首先，我们看DynamicRendezvousHandler.next_rendezvous的实现逻辑（注：ElasticAgent通过调用该接口实现的node间的协调）。DynamicRendezvousHandler.next_rendezvous 一共由5个步骤组成：
DynamicRendezvousHandler._stop_heartbeats()：停止先TCPStore的心跳操作，通过调用定时器_PeriodicTimer的cancel接口实现；Execute Exit OP：执行退出逻辑，如果当前node已经在participants中了，则先把当前节点从_RendezvousState的participants列表中删除；Execute Join OP: 下图仅描述了一个常规的场景，源码中还有一些特殊情况需要处理；
- 将自己加入到_RendezvousState的participants列表中；
- 向TCPStore发起心跳，等待至少min个node加入；
- 当_RendezvousState的participants的个数大于min时，mark rendezvous；
- 此时，Join OP执行完成，返回给_RendezvousOpExecutor 个Finish action；
DynamicRendezvousHandler._start_heartbeats()：开启心跳，这个逻辑通过_PeriodicTimer定期执行_RendezvousKeepAliveOp实现；_RendezvousKeepAliveOp的操作则是对_RendezvousState的last_heartbeats进行更新来实现；- DynamicRendezvousHandler._get_world()：从_RendezvousState中获取当前rank和work_size信息；
下面我们再看下Rendezvous的OP是如何执行的。上文提到OP是通过_DistributedRendezvousOpExecutor.run()接口统一来完成的。
- 主流程包裹在while循环中，直到OP的action为finish方可退出循环；
- 首先，会调用_BackendRendezvousStateHolder.sync()接口在所有node间进行_RendezvousState的同步；
- 若当前node有内容需要更新，则调用C10dRendezvousBackend.set_state()来更新；若没有，则调用C10dRendezvousBackend.get_state()来获取最新的state；
- 若获取了最新的state，则对当前node上存储的state进行更新；
- 然后，调用当前需要执行的OP，OP接口会返回一个ACTION，_DistributedRendezvousOpExecutor则根据ACTION的内容执行keep_alive/add_to_participants/add_to_wait_list等操作;
6 Failover
Failover分为两种情况：
- ElasticAgent Process正常，但是worker process 出错
- ElasticAgent Process 异常退出
6.1 Worker Fail
对于worker fail的场景，worker process的异常状态会被ElasticAgent捕获，实现逻辑在SimpleElasticAgent的_invoke_run接口中。
- 该接口实现中会循环monitor 当前node上所有worker process的状态，如果process 异常，则会进行入UNHEALTHY/FAILED状态的处理流程。
- 如果当前重试的次数小于_remain_restart，则会发起restart worker的流程
restart worker的实现逻辑也很清晰：
- 先stop 点前node上所有worker
- 然后重新走_initialize_workers逻辑来进行Rendezvous和start worker
def _restart_workers(self, worker_group: WorkerGroup) -> None:
"""
Restarts (stops, rendezvous, starts) all local workers in the group.
"""
role = worker_group.spec.role
log.info(f"[{role}] Stopping worker group")
self._stop_workers(worker_group)
worker_group.state = WorkerState.STOPPED
self._initialize_workers(worker_group)6.2 ElasticAgent Fail
首先，我们看下当一个node Fail掉后，弹性训练是如何运行的。这有两个node：node0和node1，开始node0和node1同时进行分布式训练，当训练到一定时间后，我们将node1 kill掉。
这是node1上的日志：
[763] epoch 14 (rank = 4, local_rank = 0) loss = 1.2388396263122559
[765] epoch 14 (rank = 6, local_rank = 2) loss = 1.4543075561523438
[766] epoch 14 (rank = 7, local_rank = 3) loss = 1.0290627479553223
[764] epoch 14 (rank = 5, local_rank = 1) loss = 1.1143463850021362
^CTraceback (most recent call last):
Traceback (most recent call last):
File "/opt/conda/bin/torchrun", line 33, in <module>
sys.exit(load_entry_point('torch==1.11.0', 'console_scripts', 'torchrun')())
File "/opt/conda/lib/python3.8/site-packages/torch/distributed/elastic/multiprocessing/errors/__init__.py", line 345, in wrapper
return f(*args, **kwargs)
File "/opt/conda/lib/python3.8/site-packages/torch/distributed/run.py", line 724, in main
run(args)
File "/opt/conda/lib/python3.8/site-packages/torch/distributed/run.py", line 715, in run
elastic_launch(
File "/opt/conda/lib/python3.8/site-packages/torch/distributed/launcher/api.py", line 131, in __call__
return launch_agent(self._config, self._entrypoint, list(args))
File "/opt/conda/lib/python3.8/site-packages/torch/distributed/launcher/api.py", line 236, in launch_agent
result = agent.run()
File "/opt/conda/lib/python3.8/site-packages/torch/distributed/elastic/metrics/api.py", line 125, in wrapper
result = f(*args, **kwargs)
File "/opt/conda/lib/python3.8/site-packages/torch/distributed/elastic/agent/server/api.py", line 709, in run
result = self._invoke_run(role)
File "/opt/conda/lib/python3.8/site-packages/torch/distributed/elastic/agent/server/api.py", line 850, in _invoke_run
time.sleep(monitor_interval)
File "/opt/conda/lib/python3.8/site-packages/torch/distributed/elastic/multiprocessing/api.py", line 60, in _terminate_process_handler
raise SignalException(f"Process {os.getpid()} got signal: {sigval}", sigval=sigval)
torch.distributed.elastic.multiprocessing.api.SignalException: Process 759 got signal: 2这是node0上的日志，我们可以得出以下结论：
- 当Elastic Agent退出时，会导致其他存活的Elastic Agent中的process 运行失败；这是因为剩余process无法在正常进行collective communication了；
- 存活的Elastic Agent会按照UNHEALTHY/FAILED的处理逻辑来重启本机的worker；若失败的Elastic Agent没有重启，则剩余的Elastic Agent重新构建worker group继续进行训练，若失败的Elastic Agent重新启动（例如kubernetes中job提供重启的机制），则会重新加入到整个训练任务中；
# 1) 此时node0和node1共同进行分布式训练
...
[11762] epoch 14 (rank = 2, local_rank = 2) loss = 1.1763713359832764 [702/1958]
[11760] epoch 14 (rank = 0, local_rank = 0) loss = 1.324049949645996
# 2） 此时node1被kill掉，因此当执行collective communication时，会报出异常
[E ProcessGroupNCCL.cpp:406] Some NCCL operations have failed or timed out. Due to the asynchronous nature of CUDA kernels, subsequent GPU operations might run on corrupted/incomplete d
ata. To avoid this inconsistency, we are taking the entire process down.
terminate called after throwing an instance of 'std::runtime_error'
what(): NCCL error: unhandled system error, NCCL version 21.0.3
ncclSystemError: System call (socket, malloc, munmap, etc) failed.
# 3）stop 其他三个process
WARNING:torch.distributed.elastic.multiprocessing.api:Sending process 11761 closing signal SIGTERM
WARNING:torch.distributed.elastic.multiprocessing.api:Sending process 11762 closing signal SIGTERM
WARNING:torch.distributed.elastic.multiprocessing.api:Sending process 11763 closing signal SIGTERM
ERROR:torch.distributed.elastic.multiprocessing.api:failed (exitcode: -6) local_rank: 0 (pid: 11760) of binary: /opt/conda/bin/python
# 4）重新走_initialize_workers逻辑
[11828] Initializing process group with: {'MASTER_ADDR': 'iZ2ze9q3ftqtxtqlkrk6tuZ', 'MASTER_PORT': '40539', 'WORLD_SIZE': '4', 'LOCAL_WORLD_SIZE': '4'}[11825] Initializing process group
with: {'MASTER_ADDR': 'iZ2ze9q3ftqtxtqlkrk6tuZ', 'MASTER_PORT': '40539', 'WORLD_SIZE': '4', 'LOCAL_WORLD_SIZE': '4'}
[11826] Initializing process group with: {'MASTER_ADDR': 'iZ2ze9q3ftqtxtqlkrk6tuZ', 'MASTER_PORT': '40539', 'WORLD_SIZE': '4', 'LOCAL_WORLD_SIZE': '4'}
[11827] Initializing process group with: {'MASTER_ADDR': 'iZ2ze9q3ftqtxtqlkrk6tuZ', 'MASTER_PORT': '40539', 'WORLD_SIZE': '4', 'LOCAL_WORLD_SIZE': '4'}
[11827] (rank = 2, local_rank = 2) train worker starting...
[11828] (rank = 3, local_rank = 3) train worker starting...
[11825] (rank = 0, local_rank = 0) train worker starting...
[11826] (rank = 1, local_rank = 1) train worker starting...
# 5）node0 独自进行分布式训练
load checkpoint from checkpoint.ptload checkpoint from checkpoint.ptload checkpoint from checkpoint.ptload checkpoint from checkpoint.pt
[11826] epoch 14 (rank = 1, local_rank = 1) loss = 0.839302122592926
[11828] epoch 14 (rank = 3, local_rank = 3) loss = 0.8971960544586182
[11825] epoch 14 (rank = 0, local_rank = 0) loss = 1.33822691440582287 Scale Up/Down
Scale Down的可以理解为上文中Elastic Agent退出，但是没有重启的场景，因此这里不再赘述。
Scale UP这里要再介绍一下，Scale UP的流程仍旧可以用上图进行描述：
- 当有新的节点加入时，由于当前Elastic已经建立一个的Rendezvous，其无法加入，所以当前Node会被加入到_RendezvousState的wait_list中
- 当ElasticAgent和对应的worker process都正常运行时，monitor会返回Healthy的状态；此时，ElasticAgent会检查_RendezvousState的waiting list的node个数，发现waiting list大于0，则出发restart worker来发起新一轮的Rendezvous以将新的加入，这样新的Node加入到了worker group中；
二 \ 代码----
著名物理学家，诺贝尔奖得主Richard Feynman办公室的黑板上写了："What I cannot create, I do not understand."。在程序员界也经常有"show me the code"的口号。因此，我打算写一系列的分布式训练的文章，将以往抽象的分布式训练的概念以代码的形式展现出来，并保证每个代码可执行、可验证、可复现，并贡献出来源码让大家相互交流。
经过调研发现pytorch对于分布式训练做好很好的抽象且接口完善，因此本系列文章将以pytorch为主要框架进行，文章中的例子很多都来自pytorch的文档，并在此基础上进行了调试和扩充。
最后，由于分布式训练的理论介绍网络上已经很多了，理论部分的介绍不会是本系列文章的重点，我会将重点放在代码层面的介绍上面。
Pytorch - 分布式训练极简体验：https://zhuanlan.zhihu.com/p/477073906
Pytorch - 分布式通信原语（附源码）：https://zhuanlan.zhihu.com/p/478953028
Pytorch - 手写allreduce分布式训练（附源码）：https://zhuanlan.zhihu.com/p/482557067
Pytorch - 算子间并行极简实现（附源码）：https://zhuanlan.zhihu.com/p/483640235
Pytorch - 多机多卡极简实现（附源码）：https://zhuanlan.zhihu.com/p/486130584
1. 介绍
Pytorch在1.9.0引入了torchrun，用其替代1.9.0以前版本的torch.distributed.launch。torchrun在torch.distributed.launch 功能的基础上主要新增了两个功能：
- Failover: 当worker训练失败时，会自动重新启动所有worker继续进行训练；
- Elastic: 可以动态增加或或删除node节点，本文将通过一个例子说明Elastic Training应该如何使用；
本例中会先在Node0上启动4 GPU的worker group ，等其训练一段时间后，会在Node1上再启动4 GPU的workers，并与Node1上的workers构成一个新的worker group，最终构成一个2机8卡的分布式训练。
2. 模型构建
一个简单的全连接模型神经网络模型
class ToyModel(nn.Module):
def __init__(self):
super(ToyModel, self).__init__()
self.net1 = nn.Linear(10, 10)
self.relu = nn.ReLU()
self.net2 = nn.Linear(10, 5)
def forward(self, x):
return self.net2(self.relu(self.net1(x)))3. checkpoint 处理
由于再每次增加或删除node时，会将所有worker kill掉，然后再重新启动所有worker进行训练。因此，在训练代码中要对训练的状态进行保存，以保证重启后能接着上次的状态继续训练。
需要保存的信息一般有如下内容：
- model ：模型的参数信息
- optimizer ：优化器的参数信心
- epoch：当前执行到第几个epoch
save和load的代码如下所示
torch.save：利用python的pickle将python的object 进行序列化，并保存到本地文件；torch.load: 将torch.save后的本地文件进行反序列化，并加载到内存中；model.state_dict():存储了model 每个layer和其对应的param信息optimizer.state_dict()：存储了优化器的参数信信息
def save_checkpoint(epoch, model, optimizer, path):
torch.save({
"epoch": epoch,
"model_state_dict": model.state_dict(),
"optimize_state_dict": optimizer.state_dict(),
}, path)
def load_checkpoint(path):
checkpoint = torch.load(path)
return checkpoint4. 训练代码
初始化逻辑如下：
- 1~3行: 输出当前worker的关键环境变量，用于后面的结果展示
- 5~8行：创建模型、优化器和损失函数
- 10~12行：初始化参数信息
- 14~19行：如果存在checkpoint，则加载checkpoint，并赋值给model、optimizer和firt_epoch
local_rank = int(os.environ["LOCAL_RANK"])
rank = int(os.environ["RANK"])
print(f"[{os.getpid()}] (rank = {rank}, local_rank = {local_rank}) train worker starting...")
model = ToyModel().cuda(local_rank)
ddp_model = DDP(model, [local_rank])
loss_fn = nn.MSELoss()
optimizer = optim.SGD(ddp_model.parameters(), lr=0.001)
optimizer.zero_grad()
max_epoch = 100
first_epoch = 0
ckp_path = "checkpoint.pt"
if os.path.exists(ckp_path):
print(f"load checkpoint from {ckp_path}")
checkpoint = load_checkpoint(ckp_path)
model.load_state_dict(checkpoint["model_state_dict"])
optimizer.load_state_dict(checkpoint["optimize_state_dict"])
first_epoch = checkpoint["epoch"]训练逻辑：
- 1行：epoch执行的次数为first_epoch到max_epoch，以便能够在worker被重启后继续原有的epoch继续训练；
- 2行：为了展示动态添加node效果，这里添加sleep函数来降低训练的速度；
- 3~8行：模型训练流程；
- 9行：为了简单，文本每个epoch进行一次checkpoint保存；将当前的epoch，model和optimizer保存到checkpoint中；
for i in range(first_epoch, max_epoch):
time.sleep(1) # 为了展示动态添加node效果，这里添加sleep函数来降低训练的速度
outputs = ddp_model(torch.randn(20, 10).to(local_rank))
labels = torch.randn(20, 5).to(local_rank)
loss = loss_fn(outputs, labels)
loss.backward()
print(f"[{os.getpid()}] epoch {i} (rank = {rank}, local_rank = {local_rank}) loss = {loss.item()}\n")
optimizer.step()
save_checkpoint(i, model, optimizer, ckp_path)5. 启动方式
由于我们使用torchrun来启动多机多卡任务，无需使用spawn接口来启动多个进程（torchrun会负责将我们的python script启动为一个process），因此直接调用上文编写的train函数，并在前后分别添加DistributedDataParallel的初始化和效果函数即可。
下面代码描述了上文train接口的调用。
def run():
env_dict = {
key: os.environ[key]
for key in ("MASTER_ADDR", "MASTER_PORT", "WORLD_SIZE", "LOCAL_WORLD_SIZE")
}
print(f"[{os.getpid()}] Initializing process group with: {env_dict}")
dist.init_process_group(backend="nccl")
train()
dist.destroy_process_group()
if __name__ == "__main__":
run()本例中使用torchrun来执行多机多卡的分布式训练任务（注：torch.distributed.launch已经被pytorch淘汰了，尽量不要再使用）。启动脚本描述如下（注：node0和node1均通过该脚本进行启动）
--nnodes=1:3：表示当前训练任务接受最少1个node，最多3个node参与分布式训练；--nproc_per_node=4:表示每个node上节点有4个process--max_restarts=3: worker group最大的重启次数；这里需要注意的是，node fail、node scale down和node scale up都会导致restart；--rdzv_id=1：一个unique的job id，所有node均使用同一个job id；--rdzv_backend: rendezvous的backend实现，默认支持c10d和etcd两种；rendezvous用于多个node之间的通信和协调；--rdzv_endpoint：rendezvous的地址，应该为一个node的host ip和port；
torchrun \
--nnodes=1:3\
--nproc_per_node=4\
--max_restarts=3\
--rdzv_id=1\
--rdzv_backend=c10d\
--rdzv_endpoint="192.0.0.1:1234"\
train_elastic.py6. 结果分析
代码：BetterDL - train_elastic.py：https://github.com/tingshua-yts/BetterDL/blob/master/test/pytorch/DDP/train_elastic.py
运行环境: 2台4卡 v100机器
image: pytorch/pytorch:1.11.0-cuda11.3-cudnn8-runtime
gpu: v100先在node0上执行执行启动脚本
torchrun \
--nnodes=1:3\
--nproc_per_node=4\
--max_restarts=3\
--rdzv_id=1\
--rdzv_backend=c10d\
--rdzv_endpoint="192.0.0.1:1234"\
train_elastic.py得到如下结果
- 2~5行：当前启动的是单机4卡的训练任务，因此WORLD_SIZE为4， LOCAL_WORKD_SIZE也为4
- 6~9行：共有4个rank参与了分布式训练，rank0~rank3
- 10~18行: rank0~rank3 均从epoch=0开始训练
r/workspace/DDP# sh run_elastic.sh
[4031] Initializing process group with: {'MASTER_ADDR': '192.0.0.1', 'MASTER_PORT': '44901', 'WORLD_SIZE': '4', 'LOCAL_WORLD_SIZE': '4'}
[4029] Initializing process group with: {'MASTER_ADDR': '192.0.0.1', 'MASTER_PORT': '44901', 'WORLD_SIZE': '4', 'LOCAL_WORLD_SIZE': '4'}
[4030] Initializing process group with: {'MASTER_ADDR': '192.0.0.1', 'MASTER_PORT': '44901', 'WORLD_SIZE': '4', 'LOCAL_WORLD_SIZE': '4'}
[4032] Initializing process group with: {'MASTER_ADDR': '192.0.0.1', 'MASTER_PORT': '44901', 'WORLD_SIZE': '4', 'LOCAL_WORLD_SIZE': '4'}
[4029] (rank = 0, local_rank = 0) train worker starting...
[4030] (rank = 1, local_rank = 1) train worker starting...
[4032] (rank = 3, local_rank = 3) train worker starting...
[4031] (rank = 2, local_rank = 2) train worker starting...
[4101] epoch 0 (rank = 1, local_rank = 1) loss = 0.9288564920425415
[4103] epoch 0 (rank = 3, local_rank = 3) loss = 0.9711472988128662
[4102] epoch 0 (rank = 2, local_rank = 2) loss = 1.0727070569992065
[4100] epoch 0 (rank = 0, local_rank = 0) loss = 0.9402943253517151
[4100] epoch 1 (rank = 0, local_rank = 0) loss = 1.0327017307281494
[4101] epoch 1 (rank = 1, local_rank = 1) loss = 1.4485043287277222
[4103] epoch 1 (rank = 3, local_rank = 3) loss = 1.0959293842315674
[4102] epoch 1 (rank = 2, local_rank = 2) loss = 1.0669530630111694
...在node1上执行与上面相同的脚本
torchrun \
--nnodes=1:3\
--nproc_per_node=4\
--max_restarts=3\
--rdzv_id=1\
--rdzv_backend=c10d\
--rdzv_endpoint="192.0.0.1:1234"\
train_elastic.pynode1上结果如下：
- 2~5行：由于添加node1，当前执行的是2机8卡的分布式训练任务，因此WORLD_SIZE=8， LOCAL_WORLD_SIZE=4
- 6~9行：当前node1上workers的rank为rank4 ~rank7
- 13~20行: 由于node1是在node0上work训练到epoch35的时候加入的，因此其接着epoch 35开始训练
/workspace/DDP# sh run_elastic.sh
[696] Initializing process group with: {'MASTER_ADDR': '192.0.0.1', 'MASTER_PORT': '42913', 'WORLD_SIZE': '8', 'LOCAL_WORLD_SIZE': '4'}
[697] Initializing process group with: {'MASTER_ADDR': '192.0.0.1', 'MASTER_PORT': '42913', 'WORLD_SIZE': '8', 'LOCAL_WORLD_SIZE': '4'}
[695] Initializing process group with: {'MASTER_ADDR': '192.0.0.1', 'MASTER_PORT': '42913', 'WORLD_SIZE': '8', 'LOCAL_WORLD_SIZE': '4'}
[694] Initializing process group with: {'MASTER_ADDR': '192.0.0.1', 'MASTER_PORT': '42913', 'WORLD_SIZE': '8', 'LOCAL_WORLD_SIZE': '4'}
[697] (rank = 7, local_rank = 3) train worker starting...
[695] (rank = 5, local_rank = 1) train worker starting...
[694] (rank = 4, local_rank = 0) train worker starting...
[696] (rank = 6, local_rank = 2) train worker starting...
load checkpoint from checkpoint.ptload checkpoint from checkpoint.pt
load checkpoint from checkpoint.pt
load checkpoint from checkpoint.pt
[697] epoch 35 (rank = 7, local_rank = 3) loss = 1.1888569593429565
[694] epoch 35 (rank = 4, local_rank = 0) loss = 0.8916441202163696
[695] epoch 35 (rank = 5, local_rank = 1) loss = 1.5685604810714722
[696] epoch 35 (rank = 6, local_rank = 2) loss = 1.11683189868927
[696] epoch 36 (rank = 6, local_rank = 2) loss = 1.3724170923233032
[694] epoch 36 (rank = 4, local_rank = 0) loss = 1.061527967453003
[695] epoch 36 (rank = 5, local_rank = 1) loss = 0.96876460313797
[697] epoch 36 (rank = 7, local_rank = 3) loss = 0.8060566782951355
...node0上结果如下：
- 6~9行: node0上的works在执行到epoch 35时，node1上执行了训练脚本，请求加入到训练任务中
- 10~13行：所有workers重新启动，由于添加了node1，当前执行的是2机8卡的分布式训练任务，因此WORLD_SIZE=8， LOCAL_WORLD_SIZE=4
- 14~17行：当前node1上works的rank为rank0~rank3
- 18~21行：加载checkpoint
- 22~30行：接着checkpoint中的model、optimizer和epoch继续训练
...
[4100] epoch 35 (rank = 0, local_rank = 0) loss = 1.0746158361434937
[4101] epoch 35 (rank = 1, local_rank = 1) loss = 1.1712706089019775
[4103] epoch 35 (rank = 3, local_rank = 3) loss = 1.1774182319641113
[4102] epoch 35 (rank = 2, local_rank = 2) loss = 1.0898035764694214
WARNING:torch.distributed.elastic.multiprocessing.api:Sending process 4100 closing signal SIGTERM
WARNING:torch.distributed.elastic.multiprocessing.api:Sending process 4101 closing signal SIGTERM
WARNING:torch.distributed.elastic.multiprocessing.api:Sending process 4102 closing signal SIGTERM
WARNING:torch.distributed.elastic.multiprocessing.api:Sending process 4103 closing signal SIGTERM
[4164] Initializing process group with: {'MASTER_ADDR': '192.0.0.1', 'MASTER_PORT': '42913', 'WORLD_SIZE': '8', 'LOCAL_WORLD_SIZE': '4'}
[4165] Initializing process group with: {'MASTER_ADDR': '192.0.0.1', 'MASTER_PORT': '42913', 'WORLD_SIZE': '8', 'LOCAL_WORLD_SIZE': '4'}
[4162] Initializing process group with: {'MASTER_ADDR': '192.0.0.1', 'MASTER_PORT': '42913', 'WORLD_SIZE': '8', 'LOCAL_WORLD_SIZE': '4'}
[4163] Initializing process group with: {'MASTER_ADDR': '192.0.0.1', 'MASTER_PORT': '42913', 'WORLD_SIZE': '8', 'LOCAL_WORLD_SIZE': '4'}
[4162] (rank = 0, local_rank = 0) train worker starting...
[4163] (rank = 1, local_rank = 1) train worker starting...
[4164] (rank = 2, local_rank = 2) train worker starting...
[4165] (rank = 3, local_rank = 3) train worker starting...
load checkpoint from checkpoint.pt
load checkpoint from checkpoint.pt
load checkpoint from checkpoint.pt
load checkpoint from checkpoint.pt
[4165] epoch 35 (rank = 3, local_rank = 3) loss = 1.3437936305999756
[4162] epoch 35 (rank = 0, local_rank = 0) loss = 1.5693414211273193
[4163] epoch 35 (rank = 1, local_rank = 1) loss = 1.199862003326416
[4164] epoch 35 (rank = 2, local_rank = 2) loss = 1.0465545654296875
[4163] epoch 36 (rank = 1, local_rank = 1) loss = 0.9741991758346558
[4162] epoch 36 (rank = 0, local_rank = 0) loss = 1.3609280586242676
[4164] epoch 36 (rank = 2, local_rank = 2) loss = 0.9585908055305481
[4165] epoch 36 (rank = 3, local_rank = 3) loss = 0.9169824123382568
...
