AI4HealthUOL/SSSD

question about zeroconv

JunSeok94 opened this issue · 1 comments

Hi, I wonder why you use zeroconv instead of normal conv?

These zero line means the weight and bias of conv1d is 0. I didn't understand why this operation is needed.

self.conv.weight.data.zero_()
self.conv.bias.data.zero_()

The implementation closely follows WaveNet vocoder, see https://github.com/philsyn/DiffWave-Vocoder/blob/master/WaveNet.py