Channel_shuffle

Usage

torch_channel_shuffle(self, groups)

Arguments

self: (Tensor) the input tensor
groups: (int) number of groups to divide channels in and rearrange.

Divide the channels in a tensor of shape

math:(*, C , H, W) :

Divide the channels in a tensor of shape $(*, C, H, W)$ into g groups and rearrange them as $(*, C \frac{g}{,} g, H, W)$ , while keeping the original tensor shape.

Examples

if (torch_is_installed()) {

input <- torch_randn(c(1, 4, 2, 2))
print(input)
output <- torch_channel_shuffle(input, 2)
print(output)
}
#> torch_tensor
#> (1,1,.,.) = 
#>   3.0396 -0.4883
#>   0.2065 -0.0433
#> 
#> (1,2,.,.) = 
#>   1.6674  0.4861
#>   0.4390  0.3384
#> 
#> (1,3,.,.) = 
#>  -1.5326  0.3339
#>   1.0767 -0.7138
#> 
#> (1,4,.,.) = 
#>  -0.0964  0.0504
#>   1.4690  1.5594
#> [ CPUFloatType{1,4,2,2} ]
#> torch_tensor
#> (1,1,.,.) = 
#>   3.0396 -0.4883
#>   0.2065 -0.0433
#> 
#> (1,2,.,.) = 
#>  -1.5326  0.3339
#>   1.0767 -0.7138
#> 
#> (1,3,.,.) = 
#>   1.6674  0.4861
#>   0.4390  0.3384
#> 
#> (1,4,.,.) = 
#>  -0.0964  0.0504
#>   1.4690  1.5594
#> [ CPUFloatType{1,4,2,2} ]