Merge #1129

1129: Added dropgrad in huber_loss r=CarloLucibello a=HenriDeh

Workaround to prevent `iterate(::nothing)` when working with CuArrays. See issue #1128

Co-authored-by: HenriDeh <47037088+HenriDeh@users.noreply.github.com>

src/layers/stateless.jl

@@ -46,9 +46,10 @@ given the prediction `ŷ` and true values `y`.
     Huber loss = |
                  |  δ * (|ŷ - y| - 0.5 * δ), otherwise
 """
+#TODO: remove dropgrad when Zygote can handle this function with CuArrays
 function huber_loss(ŷ, y;  δ=eltype(ŷ)(1))
    abs_error = abs.(ŷ .- y)
-   temp = abs_error .<  δ
+   temp = Zygote.dropgrad(abs_error .<  δ)
    x = eltype(ŷ)(0.5)
    hub_loss = sum(((abs_error.^2) .* temp) .* x .+ δ*(abs_error .- x*δ) .* (1 .- temp)) * 1 // length(y)
 end