The cumprod rule with CUDA hits scalar indexing error

[Red-Portal]

Hi, it seems like the rule is not properly kicking in for CUDA.

using Zygote, CUDA
gradient(x -> x |> cumprod |> sum , CUDA.randn(3) )

ERROR: Scalar indexing is disallowed.
Invocation of getindex resulted in scalar indexing of a GPU array.
This is typically caused by calling an iterating implementation of a method.
Such implementations *do not* execute on the GPU, but very slowly on the CPU,
and therefore are only permitted from the REPL for prototyping purposes.
If you did intend to index this array, annotate the caller with @allowscalar.
Stacktrace:
  [1] error(s::String)
    @ Base ./error.jl:35
  [2] assertscalar(op::String)
    @ GPUArraysCore ~/.julia/packages/GPUArraysCore/uOYfN/src/GPUArraysCore.jl:103
  [3] getindex
    @ ~/.julia/packages/GPUArrays/5XhED/src/host/indexing.jl:9 [inlined]
  [4] ∇cumprod!
    @ ~/.julia/packages/ChainRules/9sNmB/src/rulesets/Base/mapreduce.jl:405 [inlined]
  [5] ∇cumprod(x::CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, dy::CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, y::CuArray{Float32, 1, CUDA.Mem.DeviceBuffer})
    @ ChainRules ~/.julia/packages/ChainRules/9sNmB/src/rulesets/Base/mapreduce.jl:395
  [6] #1693
    @ ~/.julia/packages/ChainRules/9sNmB/src/rulesets/Base/mapreduce.jl:341 [inlined]
  [7] unthunk
    @ ~/.julia/packages/ChainRulesCore/0t04l/src/tangent_types/thunks.jl:204 [inlined]
  [8] unthunk
    @ ~/.julia/packages/ChainRulesCore/0t04l/src/tangent_types/thunks.jl:237 [inlined]
  [9] wrap_chainrules_output
    @ ~/.julia/packages/Zygote/4rucm/src/compiler/chainrules.jl:110 [inlined]
 [10] map
    @ ./tuple.jl:274 [inlined]
 [11] wrap_chainrules_output
    @ ~/.julia/packages/Zygote/4rucm/src/compiler/chainrules.jl:111 [inlined]
 [12] ZBack
    @ ~/.julia/packages/Zygote/4rucm/src/compiler/chainrules.jl:211 [inlined]
 [13] Pullback
    @ ./operators.jl:907 [inlined]
 [14] Pullback
    @ ./REPL[7]:1 [inlined]
 [15] (::Zygote.var"#75#76"{Zygote.Pullback{Tuple{var"#31#32", CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}}, Tuple{Zygote.Pullback{Tuple{typeof(|>), CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, typeof(cumprod)}, Tuple{Zygote.ZBack{ChainRules.var"#cumprod_pullback_1#1694"{Int64, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, ProjectTo{AbstractArray, NamedTuple{(:element, :axes), Tuple{ProjectTo{Float32, NamedTuple{(), Tuple{}}}, Tuple{Base.OneTo{Int64}}}}}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}}}}}, Zygote.Pullback{Tuple{typeof(|>), CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, typeof(sum)}, Tuple{Zygote.var"#4229#back#1457"{Zygote.var"#1453#1456"{CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}}}}}}}})(Δ::Float32)
    @ Zygote ~/.julia/packages/Zygote/4rucm/src/compiler/interface.jl:45
 [16] gradient(f::Function, args::CuArray{Float32, 1, CUDA.Mem.DeviceBuffer})
    @ Zygote ~/.julia/packages/Zygote/4rucm/src/compiler/interface.jl:97
 [17] top-level scope
    @ REPL[7]:1
 [18] top-level scope
    @ ~/.julia/packages/CUDA/tVtYo/src/initialization.jl:185

[ToucheSir]

Ref. #282 (comment). Closing since the rule lives in ChainRules (Zygote never had one, I think).

[Red-Portal]

Oops sorry. I should have taken this to chainrules