am trying to train a supervised classification model using pl_bolts. I am using the resnet18 model supplied with pl_bolts. Even if doing print(model) returns
...
(1): BasicBlock(
(conv1): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
(conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
)
(avgpool): AdaptiveAvgPool2d(output_size=(1, 1))
(fc): Linear(in_features=512, out_features=2, bias=True)
)
(criteria): BCELoss()
)
doing model.forward(input).size() returns [batch_size, 512]. Diving into the resnet implementation (from here lightning-bolts/resnets.py at 8b4d90d7443ac263fd246e2745c267439f4b9274 · Lightning-AI/lightning-bolts · GitHub), the forward method looks like this :
def forward(self, x):
x0 = self.conv1(x)
x0 = self.bn1(x0)
x0 = self.relu(x0)
x0 = self.maxpool(x0)
if self.return_all_feature_maps:
x1 = self.layer1(x0)
x2 = self.layer2(x1)
x3 = self.layer3(x2)
x4 = self.layer4(x3)
return [x0, x1, x2, x3, x4]
else:
x0 = self.layer1(x0)
x0 = self.layer2(x0)
x0 = self.layer3(x0)
x0 = self.layer4(x0)
x0 = self.avgpool(x0)
x0 = torch.flatten(x0, 1)
return [x0]
it does not use the fully connected layer declared at line 196. Is there a precise reason for this ? Is this because the model is intended for use in self-supervised learning and not classification ?