Using GPU: True
Epoch 1/10
----------
---------------------------------------------------------------------------
RuntimeError Traceback (most recent call last)
<ipython-input-13-cd4c03781ec7> in <module>()
8 exp_lr_scheduler = lr_scheduler.StepLR(optimizer, step_size=5, gamma=0.1) #decay
9 # 0.1 every 5 epochs
---> 10 model_ft = train_model(net, criterion, optimizer, exp_lr_scheduler, num_epochs=10)
2 frames
<ipython-input-12-e791953bf1c6> in forward(self, x)
17 # If the size is a square you can only specify a single number
18 x = F.max_pool2d(F.relu(self.conv2(x)), 2)
---> 19 x = x.view(x.size(0), 3 * 224 * 224)
20 x = F.relu(self.fc1(x))
21 x = F.relu(self.fc2(x))
I have this problem but I don’t know how to solve this. Could someone help me, please?
That’s my Net
class Net(nn.Module):
def __init__(self):
super(Net, self).__init__()
# 1 input image channel, 6 output channels, 3x3 square convolution
# kernel
self.conv1 = nn.Conv2d(3, 6, 3)
self.conv2 = nn.Conv2d(6, 16, 3)
# an affine operation: y = Wx + b
self.fc1 = nn.Linear(3 * 224 * 224, 144) # 6*6 from image dimension
self.fc2 = nn.Linear(120, 84)
self.fc3 = nn.Linear(84, 10)
def forward(self, x):
# Max pooling over a (2, 2) window
x = F.max_pool2d(F.relu(self.conv1(x)), (2, 2))
# If the size is a square you can only specify a single number
x = F.max_pool2d(F.relu(self.conv2(x)), 2)
x = x.view(x.size(0), 3 * 224 * 224)
x = F.relu(self.fc1(x))
x = F.relu(self.fc2(x))
x = self.fc3(x)
return x
def num_flat_features(self, x):
size = x.size()[1:] # all dimensions except the batch dimension
num_features = 1
for s in size:
num_features *= s
return num_features
I expect, your problem is when using view x = x.view(x.size(0), 3 * 224 * 224) which is giving error,
without providing explicit shape it might be better to use x.view(x.shape[0],-1)
Another problem might be in self.fc1 and self.fc2
As we know, torch.nn.Linear(in_features, out_features)
so you need to change,
fc1 : in_features to be shape of self.conv2 returning shape, in my case if input is 224x224 then
and also out_features should be input of fc2 as in example,
please, could you print me your input shape.
I tried, Here is code
class Net(nn.Module):
def __init__(self):
super(Net, self).__init__()
# 1 input image channel, 6 output channels, 3x3 square convolution
# kernel
self.conv1 = nn.Conv2d(3, 6, 3)
self.conv2 = nn.Conv2d(6, 16, 3)
# an affine operation: y = Wx + b
self.fc1 = nn.Linear(46656, 144) # 6*6 from image dimension
self.fc2 = nn.Linear(144, 84)
self.fc3 = nn.Linear(84, 10)
def forward(self, x):
# Max pooling over a (2, 2) window
x = F.max_pool2d(F.relu(self.conv1(x)), (2, 2))
# If the size is a square you can only specify a single number
x = F.max_pool2d(F.relu(self.conv2(x)), 2)
x = x.view(x.shape[0], -1)
print(x.shape) # prints flatten shape
x = F.relu(self.fc1(x))
x = F.relu(self.fc2(x))
x = self.fc3(x)
return x
def num_flat_features(self, x):
size = x.size()[1:] # all dimensions except the batch dimension
num_features = 1
for s in size:
num_features *= s
return num_features
It does not throws error when you have input t = torch.randn([1, 3, 224,224]). However if input size is different it might be problem, you should use transforms in Dataloader to reduce size to applicable for model.
Sorry, you’re right! I did it wrong before reshape.
def forward(self, x):
# Max pooling over a (2, 2) window
x = F.max_pool2d(F.relu(self.conv1(x)), (2, 2))
# If the size is a square you can only specify a single number
x = F.max_pool2d(F.relu(self.conv2(x)), 2)
#print(x.shape)
x = x.view(x.shape[0], -1)
print(x.shape)
x = F.relu(self.fc1(x))
x = F.relu(self.fc2(x))
x = self.fc3(x)
return x
You can get it from just printing below reshaped shape. As torch.Size([128, 46656]) is flattened shape where 128: is no of batch size 46656 is tensor need to passed to the Linear in_features.
This only happens when you have input of shape [batch, channel, h,w] i.e [128,3,224,224] . If you wanted to have different input shape , you can see reshape shape and use it accordingly.