AttributeError: 'EfficientNet' object has no attribute 'fc'

vision
1

Hi All,

I am trying to create an image classifier using this [tutorial]. (Transfer Learning for Computer Vision Tutorial — PyTorch Tutorials 1.13.1+cu117 documentation) In my case I am trying to use the EfficientNet model. I plugged the regent model instead of the resnet model used in the tutorial. However, I get an error for the code. Please find the code below and the respective error.

#model_conv = torchvision.models.regnet_y_32gf(weights = 'RegNet_Y_32GF_Weights.IMAGENET1K_SWAG_E2E_V1')
model_conv = torchvision.models.efficientnet_b7(weights = 'EfficientNet_B7_Weights.IMAGENET1K_V1')

for param in model_conv.parameters():
    param.requires_grad = False

# Parameters of newly constructed modules have requires_grad=True by default
num_ftrs = model_conv.fc.in_features
model_conv.fc = nn.Linear(num_ftrs, 2)

model_conv = model_conv.to(device)

criterion = nn.CrossEntropyLoss()

# Observe that only parameters of final layer are being optimized as
# opposed to before.
optimizer_conv = optim.SGD(model_conv.fc.parameters(), lr=0.001, momentum=0.9)

# Decay LR by a factor of 0.1 every 7 epochs
exp_lr_scheduler = lr_scheduler.StepLR(optimizer_conv, step_size=7, gamma=0.1)

Error:

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
/tmp/ipykernel_23/272062303.py in <module>
      6 
      7 # Parameters of newly constructed modules have requires_grad=True by default
----> 8 num_ftrs = model_conv.fc.in_features
      9 model_conv.fc = nn.Linear(num_ftrs, 2)
     10 

/opt/conda/lib/python3.7/site-packages/torch/nn/modules/module.py in __getattr__(self, name)
   1264                 return modules[name]
   1265         raise AttributeError("'{}' object has no attribute '{}'".format(
-> 1266             type(self).__name__, name))
   1267 
   1268     def __setattr__(self, name: str, value: Union[Tensor, 'Module']) -> None:

AttributeError: 'EfficientNet' object has no attribute 'fc'

Would anyone be able to help me in this matter.
Thanks & Best Regards
AMJS

2

You would have to use the .classifier attribute as torchvision’s EfficientNet implementations don’t use the .fc attribute name.

3

Hi,
Thanks for the reply. However, when I replace .fc with .classifier I get the following error.

AttributeError: 'Sequential' object has no attribute 'in_features'

Would you be able to help me in this matters please.
Thanks & Best Regards
AMJS

4 
self.classifier = nn.Sequential(
            nn.Dropout(p=dropout, inplace=True),
            nn.Linear(lastconv_output_channels, num_classes),
        )

This is what it is in the code. If you want the Linear layer’s number of features, you might need to use an index of 1:

num_ftrs = model_conv.classifier[1].in_features
2 Likes