pytorch多gpu并行案例

在机器学习和深度学习领域中，PyTorch已成为众多数据科学家、研究人员和开发人员的首选框架。PyTorch具有交互式和动态图形特性，因此易于使用和调试。一方面，PyTorch也支持多GPU并行运算，可以加速训练过程并提升深度学习模型的效率与精度。

下面将介绍一些pytorch多GPU并行的案例：

1. 使用nn.DataParallel

nn.DataParallel是pytorch提供的一种简单的方式，用于在多个GPU上进行模型训练。该方法适用于需要在单个GPU上加载并训练大型模型的情景。

这里提供一个范例：利用ResNet模型对CIFAR-10数据集进行分类。可以通过以下代码来实现多GPU并行训练。

import torch
import torch.nn as nn
import torch.optim as optim
from torch.optim import lr_scheduler
import torchvision
from torchvision import datasets, models, transforms
import time
import os
import copy

device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")

if not os.path.isdir('data'):
    os.mkdir('data')

data_transforms = {
    'train': transforms.Compose([
        transforms.RandomResizedCrop(224),
        transforms.RandomHorizontalFlip(),
        transforms.ToTensor(),
        transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
    ]),
    'val': transforms.Compose([
        transforms.Resize(256),
        transforms.CenterCrop(224),
        transforms.ToTensor(),
        transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
    ]),
}

data_dir = 'data/cifar10'
image_datasets = {x: datasets.CIFAR10(data_dir, train=(x == 'train'),
                                      download=True, transform=data_transforms[x])
                  for x in ['train', 'val']}
dataloaders = {x: torch.utils.data.DataLoader(image_datasets[x], batch_size=32,
                                              shuffle=True, num_workers=4)
               for x in ['train', 'val']}
dataset_sizes = {x: len(image_datasets[x]) for x in ['train', 'val']}
class_names = image_datasets['train'].classes

def train_model(model, criterion, optimizer, scheduler, num_epochs=25):
    since = time.time()
    best_model_wts = copy.deepcopy(model.state_dict())
    best_acc = 0.0
    
    for epoch in range(num_epochs):
        print('Epoch {}/{}'.format(epoch, num_epochs - 1))
        print('-' * 10)
        
        for phase in ['train', 'val']:
            if phase == 'train':
                model.train()  # Set model to training mode
            else:
                model.eval()   # Set model to evaluate mode
                
            running_loss = 0.0
            running_corrects = 0
            
            for inputs, labels in dataloaders[phase]:
                inputs = inputs.to(device)
                labels = labels.to(device)
                
                optimizer.zero_grad()
                
                with torch.set_grad_enabled(phase == 'train'):
                    outputs = model(inputs)
                    _, preds = torch.max(outputs, 1)
                    loss = criterion(outputs, labels)
                    
                    if phase == 'train':
                        loss.backward()
                        optimizer.step()
                
                running_loss += loss.item() * inputs.size(0)
                running_corrects += torch.sum(preds == labels.data)
                
            if phase == 'train':
                scheduler.step()
                
            epoch_loss = running_loss / dataset_sizes[phase]
            epoch_acc = running_corrects.double() / dataset_sizes[phase]
            
            print('{} Loss: {:.4f} Acc: {:.4f}'.format(
                phase, epoch_loss, epoch_acc))
            
            if phase == 'val' and epoch_acc > best_acc:
                best_acc = epoch_acc
                best_model_wts = copy.deepcopy(model.state_dict())
                
        print()
        
    time_elapsed = time.time() - since
    print('Training complete in {:.0f}m {:.0f}s'.format(
        time_elapsed // 60, time_elapsed % 60))
    print('Best val Acc: {:4f}'.format(best_acc))
    
    model.load_state_dict(best_model_wts)
    return model

model_ft = models.resnet18(pretrained=True)
num_ftrs = model_ft.fc.in_features
model_ft.fc = nn.Linear(num_ftrs, 10)

model_ft = model_ft.to(device)

criterion = nn.CrossEntropyLoss()

optimizer_ft = optim.SGD(model_ft.parameters(), lr=0.001, momentum=0.9)

exp_lr_scheduler = lr_scheduler.StepLR(optimizer_ft, step_size=7, gamma=0.1)

model_ft = nn.DataParallel(model_ft)

model_ft = train_model(model_ft, criterion, optimizer_ft, exp_lr_scheduler, num_epochs=25)

2. 使用nn.parallel.DistributedDataParallel

如果我们希望加快模型训练的速度，那么就可以考虑使用nn.parallel.DistributedDataParallel。这个工具允许我们在多个GPU上宏观地将模型分发，从而更高效地进行深度学习。

使用DistributedDataParallel进行多GPU并行的方法如下：

（1）启动多个进程

首先，我们需要启动多个进程。代码如下所示：

python3 -m torch.distributed.launch --nproc_per_node=2 train.py --dist-url='tcp://127.0.0.1:8000' --world-size=2

这就会以两个进程启动主脚本。这两个进程实际上对应两个GPU，它们之间会进行通信。

（2）编写代码

在主脚本中，需要如下所示编写代码：

import torch
import torch.nn as nn
import torch.optim as optim
from torch.optim import lr_scheduler
import os
import copy

import torch.distributed as dist
import torch.multiprocessing as mp
from torch.nn.parallel import DistributedDataParallel as DDP

def setup(rank, world_size):
    os.environ['MASTER_ADDR'] = 'localhost'
    os.environ['MASTER_PORT'] = '8000'

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

def cleanup():
    dist.destroy_process_group()

class MyModel(nn.Module):
    def __init__(self):
        super(MyModel, self).__init__()
        self.conv1 = nn.Conv2d(3, 6, 5)
        self.pool = nn.MaxPool2d(2, 2)
        self.conv2 = nn.Conv2d(6, 16, 5)
        self.fc1 = nn.Linear(16 * 5 * 5, 120)
        self.fc2 = nn.Linear(120, 84)
        self.fc3 = nn.Linear(84, 10)

    def forward(self, x):
        x = self.pool(F.relu(self.conv1(x)))
        x = self.pool(F.relu(self.conv2(x)))
        x = x.view(-1, 16 * 5 * 5)
        x = F.relu(self.fc1(x))
        x = F.relu(self.fc2(x))
        x = self.fc3(x)
        return x

def train(rank, world_size):
    setup(rank, world_size)
    
    model = MyModel()
    model = DDP(model.to(rank))

    criterion = nn.CrossEntropyLoss()

    optimizer = optim.SGD(model.parameters(), lr=0.001, momentum=0.9)

    exp_lr_scheduler = lr_scheduler.StepLR(optimizer, step_size=7, gamma=0.1)

    train_loader = get_data_loader(num_replicas=world_size, rank=rank)

    for epoch in range(25):
        for batch_idx, (data, target) in enumerate(train_loader):
            optimizer.zero_grad()
            output = model(data.to(rank))
            loss = criterion(output, target.to(rank))
            loss.backward()
            optimizer.step()

    cleanup()

if __name__ == '__main__':
    mp.spawn(train, args=(2, 2), nprocs=2, join=True)

其中，setup()函数启动了进程组，并使所有进程可以互相通信。cleanup()函数在训练完之后关闭进程组。

总结起来，PyTorch的多GPU并行训练的方法较为灵活，可根据实际需求选择适用的方法。无论是使用nn.DataParallel还是nn.parallel.DistributedDataParallel，都能在多个GPU上加速模型训练。