nn_softmax2d function

Softmax2d module

Applies SoftMax over features to each spatial location. When given an image of Channels x Height x Width, it will apply Softmax to each location $(Channels, h_i, w_j)$

nn_softmax2d()

Returns

a Tensor of the same dimension and shape as the input with values in the range [0, 1]

Shape

• Input: $(N, C, H, W)$
• Output: $(N, C, H, W)$ (same shape as input)

Examples

if (torch_is_installed()) {
m <- nn_softmax2d()
input <- torch_randn(2, 3, 12, 13)
output <- m(input)
}