Upsample¶

class dragon.vm.torch.nn.Upsample(
  size=None,
  scale_factor=None,
  mode='nearest',
  align_corners=False
)[source]¶

Upsample input via interpolating neighborhoods.

Set size or scale_factor to determine the output size:

x = torch.ones((1, 2, 3, 4))
y = torch.nn.Upsample(size=6)(x)  # Size: (1, 2, 6, 6)
z = torch.nn.UpSample(scale_factor=2)(x)  # Size: (1, 2, 6, 8)

The interpolating method can be set by mode:

upsample_nn = torch.nn.Upsample(mode='nearest')
upsample_linear = torch.nn.UpSample(mode='linear')

See also

torch.nn.functional.interpolate(…)