| 55 | |
| 56 | |
| 57 | class Upsample(nn.Module): |
| 58 | def __init__(self, in_channels, with_conv): |
| 59 | super().__init__() |
| 60 | self.with_conv = with_conv |
| 61 | if self.with_conv: |
| 62 | self.conv = torch.nn.Conv2d(in_channels, |
| 63 | in_channels, |
| 64 | kernel_size=3, |
| 65 | stride=1, |
| 66 | padding=1) |
| 67 | |
| 68 | def forward(self, x): |
| 69 | x = torch.nn.functional.interpolate(x, scale_factor=2.0, mode="nearest") |
| 70 | if self.with_conv: |
| 71 | x = self.conv(x) |
| 72 | return x |
| 73 | |
| 74 | |
| 75 | class Downsample(nn.Module): |