A problem in net.py
xiefan-guo opened this issue · comments
Thanks for your awesome reproduce work!
But, I think you have a inadvertent error in net.py.
output_pre = (output - output_bias) / mask_sum + output_bias
I think it should be as follows:
output_pre = (output - output_bias) / mask_sum * (self.kernel_size * self.kernel_size * self.in_channels) + output_bias