(self, x)
| 49 | self.bias = nn.Parameter(torch.zeros(shape)) if bias else 0. |
| 50 | |
| 51 | def forward(self, x): |
| 52 | return F.normalize( |
| 53 | x, dim=(1 if self.channel_first else |
| 54 | -1)) * self.scale * self.gamma + self.bias |
| 55 | |
| 56 | |
| 57 | class Upsample(nn.Upsample): |
nothing calls this directly
no outgoing calls
no test coverage detected