move tests

test/cuda/cuda.jl

@@ -1,12 +1,12 @@
 using Flux, Test
-using Flux.CuArrays
+using Flux.CUDA
 using Flux: gpu
 
 @info "Testing GPU Support"
 
-@testset "CuArrays" begin
+@testset "CUDA" begin
 
-CuArrays.allowscalar(false)
+CUDA.allowscalar(false)
 
 x = randn(5, 5)
 cx = gpu(x)
@@ -65,7 +65,7 @@ end
   @test gradient(foo, cu(rand(1)))[1] isa CuArray
 end
 
-if CuArrays.has_cudnn()
+if CUDA.has_cudnn()
   @info "Testing Flux/CUDNN"
   include("cudnn.jl")
   include("curnn.jl")

test/cuda/cudnn.jl

@@ -1,4 +1,4 @@
-using Flux, CuArrays, Test
+using Flux, CUDA, Test
 using Flux: pullback
 
 @testset "CUDNN BatchNorm" begin

test/cuda/curnn.jl

@@ -1,4 +1,4 @@
-using Flux, CuArrays, Test
+using Flux, CUDA, Test
 using Flux: pullback
 
 @testset for R in [RNN, GRU, LSTM]

test/cuda/layers.jl

@@ -35,7 +35,7 @@ function gradtest(name::String, layers::Vector, xs = nothing, args...)
 
           # Handle pooling layers
           if !isempty(ps)
-            @test gs[first(ps)] isa Flux.CuArrays.CuArray
+            @test gs[first(ps)] isa Flux.CUDA.CuArray
           end
         end
       end