Skip to content

Commit

Permalink
Fix Distributed Data Samplers in PyTorch Examples (kubeflow#2012)
Browse files Browse the repository at this point in the history 
Signed-off-by: Andrey Velichkevich <[email protected]>
Signed-off-by: deepanker13 <[email protected]>
  • Loading branch information
@andreyvelich @deepanker13
andreyvelich authored and deepanker13 committed Apr 8, 2024
1 parent 2205cca commit 13d4bda
Show file tree
Hide file tree
Showing 2 changed files with 436 additions and 404 deletions.
235 changes: 155 additions & 80 deletions examples/pytorch/mnist/mnist.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -10,141 +10,216 @@
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim

WORLD_SIZE = int(os.environ.get('WORLD_SIZE', 1))
from torch.utils.data import DistributedSampler


class Net(nn.Module):
def __init__(self):
super(Net, self).__init__()
self.conv1 = nn.Conv2d(1, 20, 5, 1)
self.conv2 = nn.Conv2d(20, 50, 5, 1)
self.fc1 = nn.Linear(4*4*50, 500)
self.fc1 = nn.Linear(4 * 4 * 50, 500)
self.fc2 = nn.Linear(500, 10)

def forward(self, x):
x = F.relu(self.conv1(x))
x = F.max_pool2d(x, 2, 2)
x = F.relu(self.conv2(x))
x = F.max_pool2d(x, 2, 2)
x = x.view(-1, 4*4*50)
x = x.view(-1, 4 * 4 * 50)
x = F.relu(self.fc1(x))
x = self.fc2(x)
return F.log_softmax(x, dim=1)

def train(args, model, device, train_loader, optimizer, epoch, writer):


def train(args, model, device, train_loader, epoch, writer):
model.train()
optimizer = optim.SGD(model.parameters(), lr=args.lr, momentum=args.momentum)

for batch_idx, (data, target) in enumerate(train_loader):
# Attach tensors to the device.
data, target = data.to(device), target.to(device)

optimizer.zero_grad()
output = model(data)
loss = F.nll_loss(output, target)
loss.backward()
optimizer.step()
if batch_idx % args.log_interval == 0:
print('Train Epoch: {} [{}/{} ({:.0f}%)]\tloss={:.4f}'.format(
epoch, batch_idx * len(data), len(train_loader.dataset),
100. * batch_idx / len(train_loader), loss.item()))
print(
"Train Epoch: {} [{}/{} ({:.0f}%)]\tloss={:.4f}".format(
epoch,
batch_idx * len(data),
len(train_loader.dataset),
100.0 * batch_idx / len(train_loader),
loss.item(),
)
)
niter = epoch * len(train_loader) + batch_idx
writer.add_scalar('loss', loss.item(), niter)
writer.add_scalar("loss", loss.item(), niter)


def test(args, model, device, test_loader, writer, epoch):
def test(model, device, test_loader, writer, epoch):
model.eval()
test_loss = 0

correct = 0
with torch.no_grad():
for data, target in test_loader:
# Attach tensors to the device.
data, target = data.to(device), target.to(device)

output = model(data)
test_loss += F.nll_loss(output, target, reduction='sum').item() # sum up batch loss
pred = output.max(1, keepdim=True)[1] # get the index of the max log-probability
# Get the index of the max log-probability.
pred = output.max(1, keepdim=True)[1]
correct += pred.eq(target.view_as(pred)).sum().item()

test_loss /= len(test_loader.dataset)
print('\naccuracy={:.4f}\n'.format(float(correct) / len(test_loader.dataset)))
writer.add_scalar('accuracy', float(correct) / len(test_loader.dataset), epoch)


def should_distribute():
return dist.is_available() and WORLD_SIZE > 1


def is_distributed():
return dist.is_available() and dist.is_initialized()
print("\naccuracy={:.4f}\n".format(float(correct) / len(test_loader.dataset)))
writer.add_scalar("accuracy", float(correct) / len(test_loader.dataset), epoch)


def main():
# Training settings
parser = argparse.ArgumentParser(description='PyTorch MNIST Example')
parser.add_argument('--batch-size', type=int, default=64, metavar='N',
help='input batch size for training (default: 64)')
parser.add_argument('--test-batch-size', type=int, default=1000, metavar='N',
help='input batch size for testing (default: 1000)')
parser.add_argument('--epochs', type=int, default=1, 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('--no-cuda', action='store_true', default=False,
help='disables CUDA training')
parser.add_argument('--seed', type=int, default=1, metavar='S',
help='random seed (default: 1)')
parser.add_argument('--log-interval', type=int, default=10, metavar='N',
help='how many batches to wait before logging training status')
parser.add_argument('--save-model', action='store_true', default=False,
help='For Saving the current Model')
parser.add_argument('--dir', default='logs', metavar='L',
help='directory where summary logs are stored')
if dist.is_available():
parser.add_argument('--backend', type=str, help='Distributed backend',
choices=[dist.Backend.GLOO, dist.Backend.NCCL, dist.Backend.MPI],
default=dist.Backend.GLOO)
parser = argparse.ArgumentParser(description="PyTorch FashionMNIST Example")
parser.add_argument(
"--batch-size",
type=int,
default=64,
metavar="N",
help="input batch size for training (default: 64)",
)
parser.add_argument(
"--test-batch-size",
type=int,
default=1000,
metavar="N",
help="input batch size for testing (default: 1000)",
)
parser.add_argument(
"--epochs",
type=int,
default=1,
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(
"--no-cuda",
action="store_true",
default=False,
help="disables CUDA training",
)
parser.add_argument(
"--seed",
type=int,
default=1,
metavar="S",
help="random seed (default: 1)",
)
parser.add_argument(
"--log-interval",
type=int,
default=10,
metavar="N",
help="how many batches to wait before logging training status",
)
parser.add_argument(
"--save-model",
action="store_true",
default=False,
help="For Saving the current Model",
)
parser.add_argument(
"--dir",
default="logs",
metavar="L",
help="directory where summary logs are stored",
)

parser.add_argument(
"--backend",
type=str,
help="Distributed backend",
choices=[dist.Backend.GLOO, dist.Backend.NCCL, dist.Backend.MPI],
default=dist.Backend.GLOO,
)

args = parser.parse_args()
use_cuda = not args.no_cuda and torch.cuda.is_available()
if use_cuda:
print('Using CUDA')
print("Using CUDA")
if args.backend != dist.Backend.NCCL:
print(
"Warning. Please use `nccl` distributed backend for the best performance using GPUs"
)

writer = SummaryWriter(args.dir)

torch.manual_seed(args.seed)

device = torch.device("cuda" if use_cuda else "cpu")

if should_distribute():
print('Using distributed PyTorch with {} backend'.format(args.backend))
dist.init_process_group(backend=args.backend)
# Attach model to the device.
model = Net().to(device)

kwargs = {'num_workers': 1, 'pin_memory': True} if use_cuda else {}
print("Using distributed PyTorch with {} backend".format(args.backend))
# Set distributed training environment variables to run this training script locally.
if "WORLD_SIZE" not in os.environ:
os.environ["RANK"] = "0"
os.environ["WORLD_SIZE"] = "1"
os.environ["MASTER_ADDR"] = "localhost"
os.environ["MASTER_PORT"] = "1234"

print(f"World Size: {os.environ['WORLD_SIZE']}. Rank: {os.environ['RANK']}")

dist.init_process_group(backend=args.backend)
Distributor = nn.parallel.DistributedDataParallel
model = Distributor(model)

# Get FashionMNIST train and test dataset.
train_ds = datasets.FashionMNIST(
"../data",
train=True,
download=True,
transform=transforms.Compose([transforms.ToTensor()]),
)
test_ds = datasets.FashionMNIST(
"../data",
train=False,
download=True,
transform=transforms.Compose([transforms.ToTensor()]),
)
# Add train and test loaders.
train_loader = torch.utils.data.DataLoader(
datasets.FashionMNIST('../data', train=True, download=True,
transform=transforms.Compose([
transforms.ToTensor(),
transforms.Normalize((0.1307,), (0.3081,))
])),
batch_size=args.batch_size, shuffle=True, **kwargs)
train_ds,
batch_size=args.batch_size,
sampler=DistributedSampler(train_ds),
)
test_loader = torch.utils.data.DataLoader(
datasets.FashionMNIST('../data', train=False, transform=transforms.Compose([
transforms.ToTensor(),
transforms.Normalize((0.1307,), (0.3081,))
])),
batch_size=args.test_batch_size, shuffle=False, **kwargs)
test_ds,
batch_size=args.test_batch_size,
sampler=DistributedSampler(test_ds),
)

model = Net().to(device)
for epoch in range(1, args.epochs + 1):
train(args, model, device, train_loader, epoch, writer)
test(model, device, test_loader, writer, epoch)

if is_distributed():
Distributor = nn.parallel.DistributedDataParallel if use_cuda \
else nn.parallel.DistributedDataParallelCPU
model = Distributor(model)
if args.save_model:
torch.save(model.state_dict(), "mnist_cnn.pt")

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

for epoch in range(1, args.epochs + 1):
train(args, model, device, train_loader, optimizer, epoch, writer)
test(args, model, device, test_loader, writer, epoch)

if (args.save_model):
torch.save(model.state_dict(),"mnist_cnn.pt")

if __name__ == '__main__':
if __name__ == "__main__":
main()
Loading

0 comments on commit 13d4bda

Please sign in to comment.