Hi I have included an initialisation part in my code which looks like this:
> for m in self.modules(): if isinstance(m, (nn.Conv2d, nn.Conv3d)): nn.init.kaiming_normal_(m.weight, mode='fan_out', nonlinearity='relu') elif isinstance(m, (nn.BatchNorm2d, nn.BatchNorm3d)): nn.init.constant_(m.weight, 1) nn.init.constant_(m.bias, 0)
And I have tested my training with two different learning rates: (1) 0.0005 and (2) 0.01 but both give quite similar results and the loss is dropping at relatively same speed. I have also fixed the random values using the same seed.
Could it be because of the code I included above or other bugs in my code that cause this to happen?