I know that I could define an skip layer in Flux, like below:
struct Skip
inner
end
(m::Skip) = x .+ m.inner(x)
But how could I multiply m.inner(x) by an sparse weight mask? How could I define the layer?
I know that I could define an skip layer in Flux, like below:
struct Skip
inner
end
(m::Skip) = x .+ m.inner(x)
But how could I multiply m.inner(x) by an sparse weight mask? How could I define the layer?