I am training a binary classifier using Pytorch on Jupyter Notebook. The following is the architecture:
class AlexNet(nn.Module):
def __init__(self, num_classes=1):
super(AlexNet, self).__init__()
self.conv_base = nn.Sequential(
nn.Conv2d(3, 96, kernel_size=11, stride=4, padding=2, bias=False),
nn.BatchNorm2d(96),
nn.ReLU(inplace=True),
nn.MaxPool2d(kernel_size=3, stride=2),
nn.Conv2d(96, 256, kernel_size=5, stride=1, padding=2, bias=False),
nn.BatchNorm2d(256),
nn.ReLU(inplace=True),
nn.MaxPool2d(kernel_size=3, stride=2),
nn.Conv2d(256, 384, kernel_size=3, stride=1, padding=1),
nn.ReLU(inplace=True),
nn.Conv2d(384, 384, kernel_size=3, stride=1, padding=1),
nn.ReLU(inplace=True),
nn.Conv2d(384, 256, kernel_size=3, stride=1, padding=1),
nn.ReLU(inplace=True),
nn.MaxPool2d(kernel_size=3, stride=2),
)
self.fc_base = nn.Sequential(
nn.Dropout(),
nn.Linear(256*6*6, 4096),
nn.ReLU(inplace=True),
nn.Dropout(),
nn.Linear(4096, 4096),
nn.ReLU(inplace=True),
nn.Linear(4096, num_classes),
)
def forward(self, x):
x = self.conv_base(x)
x = x.view(x.size(0), 256*6*6)
x = self.fc_base(x)
return x
These are my parameters for training:
criterion = nn.BCELoss()
# specify optimizer
optimizer = torch.optim.SGD(model.parameters(), lr=2e-5, momentum=0.9)
When I start the training, I get the following error:
RuntimeError: after reduction step 2: device-side assert triggered
Some weird things are also happening:
-
It was training at one point in time, but the learning rate was small. So when I changed it and started the training process again, I got the error.
-
I’m also getting this error when I run this command:
model = AlexNet()
model.cuda()
the error:
RuntimeError: CUDA error: device-side assert triggered
Some solutions that I found which worked temporarily are:
- Add
os.environ['CUDA_LAUNCH_BLOCKING'] = '1'
to the notebook. 2) GPU memory being used. So I removed the process that was running the notebook.
However, nothing seems to be working permanently. Any suggestions?