hello, thank you for pytorch
I am studying beginner tutorials
when I run cifar10_tutorial.py in Deep Learning with PyTorch: A 60 minutes Blitz, I find memory leak in loss.backward()
To run using gpu and train a larger network, I revised the tutorial code like below
origianal code--------------------------------------------------------------------------------------
from torch.autograd import Variable
import torch.nn as nn
import torch.nn.functional as F
class Net(nn.Module):
def init(self):
super(Net, 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
net = Net()
import torch.optim as optim
criterion = nn.CrossEntropyLoss()
optimizer = optim.SGD(net.parameters(), lr=0.001, momentum=0.9)
for epoch in range(2): # loop over the dataset multiple times
running_loss = 0.0 for i, data in enumerate(trainloader, 0): # get the inputs inputs, labels = data
# wrap them in Variable inputs, labels = Variable(inputs), Variable(labels)
# zero the parameter gradients optimizer.zero_grad()
# forward + backward + optimize outputs = net(inputs) loss = criterion(outputs, labels) loss.backward() optimizer.step()
# print statistics running_loss += loss.data[0] if i % 2000 == 1999: # print every 2000 mini-batches print('[%d, %5d] loss: %.3f' % (epoch+1, i+1, running_loss / 2000)) running_loss = 0.0
revised code---------------------------------------------------------------------------------------------
from torch.autograd import Variable
import torch.nn as nn
import torch.nn.functional as F
class Net(nn.Module):
def init(self):
super(Net, self).init()
self.conv1 = nn.Conv2d(3,128, 5)
self.pool = nn.MaxPool2d(2, 2)
self.conv2 = nn.Conv2d(128, 128, 5)
self.fc1 = nn.Linear(128 * 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, 128 * 5 * 5) x = F.relu(self.fc1(x)) x = F.relu(self.fc2(x)) x = self.fc3(x) return x
net = Net().cuda()
import torch.optim as optim
criterion = nn.CrossEntropyLoss().cuda()
optimizer = optim.SGD(net.parameters(), lr=0.001, momentum=0.9)
for epoch in range(2): # loop over the dataset multiple times
running_loss = 0.0 for i, data in enumerate(trainloader, 0): # get the inputs inputs, labels = data
# wrap them in Variable inputs, labels = Variable(inputs.cuda()), Variable(labels.cuda())
# zero the parameter gradients optimizer.zero_grad()
# forward + backward + optimize outputs = net(inputs) loss = criterion(outputs, labels) loss.backward() optimizer.step()
# print statistics running_loss += loss.data[0] if i % 2000 == 1999: # print every 2000 mini-batches print('[%d, %5d] loss: %.3f' % (epoch+1, i+1, running_loss / 2000)) running_loss = 0.0
when I run the above code, I found memory leak(abount 70MB) from one epoch and the next.
If I delete loss.backward, memory leak doesnโt occur
Also, If I add torch.backend.cudnn.enalbled = False and donโt delete loss.backward, memory leak doesnโt occur, but speed is slow
Install information
- install method : pip install http://download.pytorch.org/whl/cu80/torch-0.1.11.post5-cp27-none-linux_x86_64.whl.
- cudnn : 5.1
- python version : 2.7
- os : linux min 18.1
- cuda : 8.0