使用 PyTorch 和 Ray Train 运行 Horovod 分布式训练#

这个基本示例展示了如何使用 PyTorch 和 Ray Train 运行 Horovod 分布式训练。

代码示例#

import argparse
import os

import horovod.torch as hvd
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
import torch.utils.data.distributed
from filelock import FileLock
from torchvision import datasets, transforms

import ray
from ray import train
from ray.train import ScalingConfig
from ray.train.horovod import HorovodTrainer


def metric_average(val, name):
    tensor = torch.tensor(val)
    avg_tensor = hvd.allreduce(tensor, name=name)
    return avg_tensor.item()


class Net(nn.Module):
    def __init__(self):
        super(Net, self).__init__()
        self.conv1 = nn.Conv2d(1, 10, kernel_size=5)
        self.conv2 = nn.Conv2d(10, 20, kernel_size=5)
        self.conv2_drop = nn.Dropout2d()
        self.fc1 = nn.Linear(320, 50)
        self.fc2 = nn.Linear(50, 10)

    def forward(self, x):
        x = F.relu(F.max_pool2d(self.conv1(x), 2))
        x = F.relu(F.max_pool2d(self.conv2_drop(self.conv2(x)), 2))
        x = x.view(-1, 320)
        x = F.relu(self.fc1(x))
        x = F.dropout(x, training=self.training)
        x = self.fc2(x)
        return F.log_softmax(x)


def setup(config):
    data_dir = config.get("data_dir", None)
    seed = config.get("seed", 42)
    batch_size = config.get("batch_size", 64)
    use_adasum = config.get("use_adasum", False)
    lr = config.get("lr", 0.01)
    momentum = config.get("momentum", 0.5)
    use_cuda = config.get("use_cuda", False)

    # Horovod: initialize library.
    hvd.init()
    torch.manual_seed(seed)

    if use_cuda:
        # Horovod: pin GPU to local rank.
        torch.cuda.set_device(hvd.local_rank())
        torch.cuda.manual_seed(seed)

    # Horovod: limit # of CPU threads to be used per worker.
    torch.set_num_threads(1)

    kwargs = {"num_workers": 1, "pin_memory": True} if use_cuda else {}
    data_dir = data_dir or "~/data"
    with FileLock(os.path.expanduser("~/.horovod_lock")):
        train_dataset = datasets.MNIST(
            data_dir,
            train=True,
            download=True,
            transform=transforms.Compose(
                [transforms.ToTensor(), transforms.Normalize((0.1307,), (0.3081,))]
            ),
        )
    # Horovod: use DistributedSampler to partition the training data.
    train_sampler = torch.utils.data.distributed.DistributedSampler(
        train_dataset, num_replicas=hvd.size(), rank=hvd.rank()
    )
    train_loader = torch.utils.data.DataLoader(
        train_dataset, batch_size=batch_size, sampler=train_sampler, **kwargs
    )

    model = Net()

    # By default, Adasum doesn't need scaling up learning rate.
    lr_scaler = hvd.size() if not use_adasum else 1

    if use_cuda:
        # Move model to GPU.
        model.cuda()
        # If using GPU Adasum allreduce, scale learning rate by local_size.
        if use_adasum and hvd.nccl_built():
            lr_scaler = hvd.local_size()

    # Horovod: scale learning rate by lr_scaler.
    optimizer = optim.SGD(model.parameters(), lr=lr * lr_scaler, momentum=momentum)

    # Horovod: wrap optimizer with DistributedOptimizer.
    optimizer = hvd.DistributedOptimizer(
        optimizer,
        named_parameters=model.named_parameters(),
        op=hvd.Adasum if use_adasum else hvd.Average,
    )

    return model, optimizer, train_loader, train_sampler


def train_epoch(
    model, optimizer, train_sampler, train_loader, epoch, log_interval, use_cuda
):
    loss = None
    model.train()
    # Horovod: set epoch to sampler for shuffling.
    train_sampler.set_epoch(epoch)
    for batch_idx, (data, target) in enumerate(train_loader):
        if use_cuda:
            data, target = data.cuda(), target.cuda()
        optimizer.zero_grad()
        output = model(data)
        loss = F.nll_loss(output, target)
        loss.backward()
        optimizer.step()
        if batch_idx % log_interval == 0:
            # Horovod: use train_sampler to determine the number of
            # examples in this worker's partition.
            print(
                "Train Epoch: {} [{}/{} ({:.0f}%)]\tLoss: {:.6f}".format(
                    epoch,
                    batch_idx * len(data),
                    len(train_sampler),
                    100.0 * batch_idx / len(train_loader),
                    loss.item(),
                )
            )
    return loss.item() if loss else None


# Horovod function API.


def train_func(config):
    num_epochs = config.get("num_epochs", 10)
    log_interval = config.get("log_interval", 10)
    use_cuda = config.get("use_cuda", False)

    model, optimizer, train_loader, train_sampler = setup(config)

    for epoch in range(num_epochs):
        loss = train_epoch(
            model, optimizer, train_sampler, train_loader, epoch, log_interval, use_cuda
        )
        train.report(dict(loss=loss))


def main(num_workers, use_gpu, kwargs):
    trainer = HorovodTrainer(
        train_func,
        train_loop_config=kwargs,
        scaling_config=ScalingConfig(use_gpu=use_gpu, num_workers=num_workers),
    )
    results = trainer.fit()
    print(results.metrics)


# Horovod Class API.


class HorovodTrainClass:
    def __init__(self, config):
        self.log_interval = config.get("log_interval", 10)
        self.use_cuda = config.get("use_cuda", False)

        if self.use_cuda:
            torch.cuda.set_device(hvd.local_rank())

        self.model, self.optimizer, self.train_loader, self.train_sampler = setup(
            config
        )

    def train(self, epoch):
        loss = train_epoch(
            self.model,
            self.optimizer,
            self.train_sampler,
            self.train_loader,
            epoch,
            self.log_interval,
            self.use_cuda,
        )
        return loss


if __name__ == "__main__":
    # Training settings
    parser = argparse.ArgumentParser(
        description="PyTorch MNIST Example",
        formatter_class=argparse.ArgumentDefaultsHelpFormatter,
    )
    parser.add_argument(
        "--batch-size",
        type=int,
        default=64,
        metavar="N",
        help="input batch size for training (default: 64)",
    )
    parser.add_argument(
        "--num-epochs",
        type=int,
        default=5,
        metavar="N",
        help="number of epochs to train (default: 10)",
    )
    parser.add_argument(
        "--lr",
        type=float,
        default=0.01,
        metavar="LR",
        help="learning rate (default: 0.01)",
    )
    parser.add_argument(
        "--momentum",
        type=float,
        default=0.5,
        metavar="M",
        help="SGD momentum (default: 0.5)",
    )
    parser.add_argument(
        "--use-gpu", action="store_true", default=False, help="enables CUDA training"
    )
    parser.add_argument(
        "--seed", type=int, default=42, metavar="S", help="random seed (default: 42)"
    )
    parser.add_argument(
        "--log-interval",
        type=int,
        default=10,
        metavar="N",
        help="how many batches to wait before logging training status",
    )
    parser.add_argument(
        "--use-adasum",
        action="store_true",
        default=False,
        help="use adasum algorithm to do reduction",
    )
    parser.add_argument(
        "--num-workers",
        type=int,
        default=2,
        help="Number of Ray workers to use for training.",
    )
    parser.add_argument(
        "--data-dir",
        help="location of the training dataset in the local filesystem ("
        "will be downloaded if needed)",
    )
    parser.add_argument(
        "--address",
        required=False,
        type=str,
        default=None,
        help="Address of Ray cluster.",
    )

    args = parser.parse_args()

    if args.address:
        ray.init(args.address)
    else:
        ray.init()

    use_cuda = args.use_gpu if args.use_gpu is not None else False

    kwargs = {
        "data_dir": args.data_dir,
        "seed": args.seed,
        "use_cuda": use_cuda,
        "batch_size": args.batch_size,
        "use_adasum": args.use_adasum if args.use_adasum else False,
        "lr": args.lr,
        "momentum": args.momentum,
        "num_epochs": args.num_epochs,
        "log_interval": args.log_interval,
    }

    main(num_workers=args.num_workers, use_gpu=use_cuda, kwargs=kwargs)

另见#