Bug for CuArray matrix multiplication

[XingyuZhang2018]

using Test
using TensorOperations
using CUDA

@testset "A*B" for atype in [Array, CuArray]
    A = atype(randn(ComplexF64, 3, 4))
    B = atype(randn(ComplexF64, 4, 5))

    C1 = similar(A, ComplexF64, 3, 5)
    @tensor C1[a,c] = A[a,b] * B[b,c]
    C2 = A * B 
    @test C1 ≈ C2
end

for CuArray I got the error:
Got exception outside of a @test
ArgumentError: No suitable backend found for tensorcontract! and tensor types CuArray{ComplexF64, 2, CUDA.DeviceMemory}, CuArray{ComplexF64, 2, CUDA.DeviceMemory} and CuArray{ComplexF64, 2, CUDA.DeviceMemory}

[kshyatt]

Working on replicating this but did you try with using CUDA, cuTENSOR (I think you need both to trigger the extension loading, see https://github.com/QuantumKitHub/TensorOperations.jl/blob/master/Project.toml#L30)

[kshyatt]

julia> using TensorOperations, CUDA, cuTENSOR

julia> using Test

julia> @testset "A*B" for atype in [Array, CuArray]
           A = atype(randn(ComplexF64, 3, 4))
           B = atype(randn(ComplexF64, 4, 5))

           C1 = similar(A, ComplexF64, 3, 5)
           @tensor C1[a,c] = A[a,b] * B[b,c]
           C2 = A * B 
           @test C1 ≈ C2
       end
Test Summary: | Pass  Total   Time
A*B           |    1      1  23.1s
Test Summary: | Pass  Total   Time
A*B           |    1      1  28.1s
2-element Vector{Any}:
 Test.DefaultTestSet("A*B", Any[], 1, false, false, true, 1.763384369519846e9, 1.763384392635983e9, false, "REPL[4]", Random.Xoshiro(0x142ede641ea488d6, 0x1835e9ef030052b9, 0x291d9cf0440acae7, 0x8817f12a3fa63d93, 0xe30c450a39672689))
 Test.DefaultTestSet("A*B", Any[], 1, false, false, true, 1.763384393301172e9, 1.76338442139379e9, false, "REPL[4]", Random.Xoshiro(0x142ede641ea488d6, 0x1835e9ef030052b9, 0x291d9cf0440acae7, 0x8817f12a3fa63d93, 0xe30c450a39672689))

With using cuTENSOR as well, this works for me. So I think we can close.

[XingyuZhang2018]

Thanks!