Model inheritance

[robsver]

Hi everyone,
I'd like to create a new resnet50 class that inherits from the default resnet50 monai model.
I tried by using the following code but it doesn't work.
Any ideas?

class ModelParallelResNet50(resnet.ResNet):
   def __init__(self, spatial_dims=3, in_channels=1, out_channels=2, *args, **kwargs):
       super(ModelParallelResNet50, self).__init__(spatial_dims=3, in_channels=1, out_channels=2,*args, **kwargs)

Thank you, have a nice day!

[Nic-Ma]

Hi @yiheng-wang-nv ,

Could you please help share some comments about this question?

Thanks in advance.

[yiheng-wang-nv]

Hi @robsver , could you please provide more information, such as the error message? Thanks

[robsver]

Hi @Nic-Ma and @yiheng-wang-nv, thank you for replying.
I modified my code, here a basic snippet:

import monai
from monai.networks.nets import resnet50, resnet

class CustomResNet50(resnet.ResNet):
    def __init__(self, spatial_dims=3, in_channels=1, out_channels=2, *args, **kwargs):
        super().__init__(block="bottleneck", layers=[3, 4, 6, 3], block_inplanes=[64, 128, 256, 512], spatial_dims=spatial_dims, n_input_channels=in_channels, num_classes=out_channels*args, **kwargs)

model = CustomResNet50(spatial_dims=3, in_channels=1, out_channels=2).cuda()

and I get this error:

Traceback (most recent call last):
  File "/home/roberto/snippet_test.py", line 8, in <module>
    model = CustomResNet50(spatial_dims=3, in_channels=1, out_channels=2).cuda()
  File "/home/roberto/snippet_test.py", line 6, in __init__
    super().__init__(block="bottleneck", layers=[3, 4, 6, 3], block_inplanes=[64, 128, 256, 512], spatial_dims=spatial_dims, n_input_channels=in_channels, num_classes=out_channels*args, **kwargs)
  File "/home/roberto/anaconda3/envs/monai/lib/python3.10/site-packages/monai/networks/nets/resnet.py", line 242, in __init__
    self.fc = nn.Linear(block_inplanes[3] * block.expansion, num_classes) if feed_forward else None
  File "/home/roberto/.local/lib/python3.10/site-packages/torch/nn/modules/linear.py", line 96, in __init__
    self.weight = Parameter(torch.empty((out_features, in_features), **factory_kwargs))
TypeError: empty() received an invalid combination of arguments - got (tuple, dtype=NoneType, device=NoneType), but expected one of:
 * (tuple of ints size, *, tuple of names names, torch.memory_format memory_format, torch.dtype dtype, torch.layout layout, torch.device device, bool pin_memory, bool requires_grad)
 * (tuple of SymInts size, *, torch.memory_format memory_format, Tensor out, torch.dtype dtype, torch.layout layout, torch.device device, bool pin_memory, bool requires_grad)

Thanks!

[yiheng-wang-nv]

Hi @robsver , you wrote a typo:

num_classes=out_channels*args

it should be

num_classes=out_channels, *args

[robsver]

Oh, sorry for the noise, my bad! I definitely missed a simple detail!
Thanks a lot, bye!