Potential 3D bug in EfficientNet Squeeze and Excitation layer #5574
Description
efficientnet.py
if self.has_se:
self._se_adaptpool = adaptivepool_type(1)
num_squeezed_channels = max(1, int(in_channels * self.se_ratio))
self._se_reduce = conv_type(in_channels=oup, out_channels=num_squeezed_channels, kernel_size=1)
self._se_reduce_padding = _make_same_padder(self._se_reduce, [1, 1])
self._se_expand = conv_type(in_channels=num_squeezed_channels, out_channels=oup, kernel_size=1)
self._se_expand_padding = _make_same_padder(self._se_expand, [1, 1])
I believe the image_size = [1, 1] in _make_same_padder() should actually be [1, 1, 1] for the EfficientNet to support 3D. This is just by inspection, but I suspect this oversight came from extending 2D code.