Questions about mask usage in convolution

[antao97]

Hi,

Thanks for your great work!

I have some questions about the mask usage in your convolution operation. I'm wondering what is the meaning to assign conv_module.__mask__ with mask. I checked that the conv_module(x) function does not consider the conv_module.__mask__ property when operating.

dynconv/classification/dynconv/layers.py

Lines 14 to 18 in 19e4c58

	def conv1x1(conv_module, x, mask, fast=False):
	w = conv_module.weight.data
	mask.flops_per_position += w.shape[0]*w.shape[1]
	conv_module.__mask__ = mask
	return conv_module(x)

Therefore, I can't get how the masks are applied in network forward propagation, such as the basicblock in

dynconv/classification/models/resnet_util.py

Lines 66 to 70 in 19e4c58

	x = dynconv.conv3x3(self.conv1, x, None, mask_dilate)
	x = dynconv.bn_relu(self.bn1, self.relu, x, mask_dilate)
	x = dynconv.conv3x3(self.conv2, x, mask_dilate, mask)
	x = dynconv.bn_relu(self.bn2, None, x, mask)
	out = identity + dynconv.apply_mask(x, mask)

It seems that only the mask in dynconv.apply_mask(x, mask) works.