Change method to get module name

[sup3rgiu]

Right now, torchinfo is using module.__class__.__name__ to retrieve the nn.Module name which will be shown in the summary. However, every PyTorch module exposes a method _get_name(), which in the default implementation simply returns self.__class__.__name__. However, a custom layer could overwrite this method to return a custom module name, avoiding to directly overwrite self.__class__.__name__ which is not a good idea in general.

Demo:

class CustomLayer(torch.nn.Module):
    def __init__(
        self,
    ) -> None:
        super().__init__()
        #self.__class__.__name__ = self._get_name()

    def forward(self, input: torch.Tensor) -> torch.Tensor:
        return input
    
    def _get_name(self) -> str:
        return "CustomFancyName"

img = torch.randn(1, 3, 224, 224)
model = CustomLayer()
summary(model, input_size=[img.shape], dtypes=[torch.float32], depth=2)

Output:

==========================================================================================
Layer (type:depth-idx)                   Output Shape              Param #
==========================================================================================
CustomFancyName                         [1, 3, 224, 224]          --
==========================================================================================
Total params: 0
Trainable params: 0
Non-trainable params: 0
Total mult-adds (Units.MEGABYTES): 0
==========================================================================================
Input size (MB): 0.60
Forward/backward pass size (MB): 0.00
Params size (MB): 0.00
Estimated Total Size (MB): 0.60
==========================================================================================