CuArray broadcast eerror

[Moelf]

pkg versions

(v1.1) pkg> st
    Status `~/.julia/environments/v1.1/Project.toml`
  [fbb218c0] BSON v0.2.2
  [6e4b80f9] BenchmarkTools v0.4.2
  [336ed68f] CSV v0.4.3
  [3a865a2d] CuArrays v1.0.1
  [a93c6f00] DataFrames v0.17.1
  [31c24e10] Distributions v0.17.0
  [587475ba] Flux v0.8.1

minimal sample:

using Flux, Flux.Data.MNIST, Statistics
using Flux: onehotbatch, onecold, crossentropy, throttle
using Base.Iterators: repeated
using CuArrays

imgs = MNIST.images()
X = hcat(float.(reshape.(imgs, :))...) |> gpu

labels = MNIST.labels()
Y = onehotbatch(labels, 0:9) |> gpu
m = Chain(
  Dense(28^2, 32, relu),
  Dense(32, 10),
  softmax) |> gpu
accuracy(x, y) = mean(onecold(m(x)) .== onecold(y))

@show accuracy(X,Y)

Error message:

ERROR: LoadError: GPU compilation of #23(CuArrays.CuKernelState, CUDAnative.CuDeviceAr
ray{Bool,1,CUDAnative.AS.Global}, Base.Broadcast.Broadcasted{Nothing,Tuple{Base.OneTo{
Int64}},typeof(==),Tuple{Base.Broadcast.Extruded{CUDAnative.CuDeviceArray{Int64,1,CUDA
native.AS.Global},Tuple{Bool},Tuple{Int64}},Base.Broadcast.Extruded{Array{Int64,1},Tup
le{Bool},Tuple{Int64}}}}) failed
KernelError: passing and using non-bitstype argument

Argument 4 to your kernel function is of type Base.Broadcast.Broadcasted{Nothing,Tuple
{Base.OneTo{Int64}},typeof(==),Tuple{Base.Broadcast.Extruded{CUDAnative.CuDeviceArray{
Int64,1,CUDAnative.AS.Global},Tuple{Bool},Tuple{Int64}},Base.Broadcast.Extruded{Array{
Int64,1},Tuple{Bool},Tuple{Int64}}}}.
That type is not isbits, and such arguments are only allowed when they are unused by t
he kernel.

Stacktrace:
 [1] check_invocation(::CUDAnative.CompilerContext, ::LLVM.Function) at /home/akako/.j
ulia/packages/CUDAnative/PFgO3/src/compiler/validation.jl:35
 [2] compile(::CUDAnative.CompilerContext) at /home/akako/.julia/packages/CUDAnative/P
FgO3/src/compiler/driver.jl:94
 [3] #compile#109(::Bool, ::Base.Iterators.Pairs{Union{},Union{},Tuple{},NamedTuple{()
,Tuple{}}}, ::Function, ::VersionNumber, ::Any, ::Any) at /home/akako/.julia/packages/
CUDAnative/PFgO3/src/compiler/driver.jl:45
 [4] compile at /home/akako/.julia/packages/CUDAnative/PFgO3/src/compiler/driver.jl:43
 [inlined]
 [5] #compile#108(::Base.Iterators.Pairs{Union{},Union{},Tuple{},NamedTuple{(),Tuple{}
}}, ::Function, ::CUDAdrv.CuDevice, ::Function, ::Any) at /home/akako/.julia/packages/
CUDAnative/PFgO3/src/compiler/driver.jl:18
 [6] compile at /home/akako/.julia/packages/CUDAnative/PFgO3/src/compiler/driver.jl:16
 [inlined]
 [7] macro expansion at /home/akako/.julia/packages/CUDAnative/PFgO3/src/execution.jl:
269 [inlined]
 [8] #cufunction#123(::Base.Iterators.Pairs{Union{},Union{},Tuple{},NamedTuple{(),Tupl
e{}}}, ::typeof(CUDAnative.cufunction), ::getfield(GPUArrays, Symbol("##23#24")), ::Ty
pe{Tuple{CuArrays.CuKernelState,CUDAnative.CuDeviceArray{Bool,1,CUDAnative.AS.Global},
Base.Broadcast.Broadcasted{Nothing,Tuple{Base.OneTo{Int64}},typeof(==),Tuple{Base.Broa
dcast.Extruded{CUDAnative.CuDeviceArray{Int64,1,CUDAnative.AS.Global},Tuple{Bool},Tupl
e{Int64}},Base.Broadcast.Extruded{Array{Int64,1},Tuple{Bool},Tuple{Int64}}}}}}) at /ho
me/akako/.julia/packages/CUDAnative/PFgO3/src/execution.jl:240
 [9] cufunction(::Function, ::Type) at /home/akako/.julia/packages/CUDAnative/PFgO3/sr
c/execution.jl:240
 [10] macro expansion at /home/akako/.julia/packages/CUDAnative/PFgO3/src/execution.jl
:208 [inlined]
 [11] macro expansion at ./gcutils.jl:87 [inlined]
 [12] macro expansion at /home/akako/.julia/packages/CUDAnative/PFgO3/src/execution.jl
:205 [inlined]
 [13] _gpu_call(::CuArrays.CuArrayBackend, ::Function, ::CuArray{Bool,1}, ::Tuple{CuAr
ray{Bool,1},Base.Broadcast.Broadcasted{Nothing,Tuple{Base.OneTo{Int64}},typeof(==),Tup
le{Base.Broadcast.Extruded{CuArray{Int64,1},Tuple{Bool},Tuple{Int64}},Base.Broadcast.E
xtruded{Array{Int64,1},Tuple{Bool},Tuple{Int64}}}}}, ::Tuple{Tuple{Int64},Tuple{Int64}
}) at /home/akako/.julia/packages/CuArrays/qZCAt/src/gpuarray_interface.jl:59
 [14] gpu_call(::Function, ::CuArray{Bool,1}, ::Tuple{CuArray{Bool,1},Base.Broadcast.B
roadcasted{Nothing,Tuple{Base.OneTo{Int64}},typeof(==),Tuple{Base.Broadcast.Extruded{C
uArray{Int64,1},Tuple{Bool},Tuple{Int64}},Base.Broadcast.Extruded{Array{Int64,1},Tuple
{Bool},Tuple{Int64}}}}}, ::Int64) at /home/akako/.julia/packages/GPUArrays/t8tJB/src/a
bstract_gpu_interface.jl:151
 [15] gpu_call at /home/akako/.julia/packages/GPUArrays/t8tJB/src/abstract_gpu_interfa
ce.jl:128 [inlined]
 [16] copyto! at /home/akako/.julia/packages/GPUArrays/t8tJB/src/broadcast.jl:48 [inli
ned]
 [17] copyto! at ./broadcast.jl:797 [inlined]
 [18] copy(::Base.Broadcast.Broadcasted{Base.Broadcast.ArrayStyle{CuArray},Tuple{Base.
OneTo{Int64}},typeof(==),Tuple{CuArray{Int64,1},Array{Int64,1}}}) at ./broadcast.jl:77
3
 [19] materialize(::Base.Broadcast.Broadcasted{Base.Broadcast.ArrayStyle{CuArray},Noth
ing,typeof(==),Tuple{CuArray{Int64,1},Array{Int64,1}}}) at ./broadcast.jl:753
 [20] accuracy(::CuArray{Float32,2}, ::Flux.OneHotMatrix{CuArray{Flux.OneHotVector,1}}
) at /home/akako/Documents/Kaggle/model-zoo/vision/mnist/mlp.jl:15
 [21] top-level scope at show.jl:555
 [22] include at ./boot.jl:326 [inlined]
 [23] include_relative(::Module, ::String) at ./loading.jl:1038
 [24] include(::Module, ::String) at ./sysimg.jl:29
 [25] exec_options(::Base.JLOptions) at ./client.jl:267
 [26] _start() at ./client.jl:436
in expression starting at /home/akako/Documents/Kaggle/model-zoo/vision/mnist/mlp.jl:1
7

[Process exited 1]

[Moelf]

current workaround:

accuracy(x, y) = mean( (onecold(m(x)) |>cpu)  .== (onecold(y) |> cpu))

[vchuravy]

What is the type of onecold(y) or onecold(m(x)):

The error says that you have a Base.Broadcast.Extruded{Array{Int64,1}... so some part of your program is working with CPU arrays.

[KristofferC]

Ref #582, #660, #612.

[DhairyaLGandhi]

#612 works for OneHot* cases.

[Moelf]

Ref #582, #660, #612.

right, I filed this issue too because the same code was running fine on an older version of Flux.

I guess I'll just have to wait for them to merge the fix.