Fix sampling algorithm

[GiggleLiu]

Now the algorithm handles the case with marginal variables.

[mroavi]

Not all tests are passing when running the test suit in my machine. The error doesn't seem to be related to the sampling algorithm though.

Error output

```jl gradient-based tensor network solvers: Error During Test at /home/mroavi/.julia/dev/TensorInference/test/cuda.jl:6 Got exception outside of a @test This object is not a GPU array Stacktrace: [1] error(s::String) @ Base ./error.jl:35 [2] backend(::Type) @ GPUArraysCore ~/.julia/packages/GPUArraysCore/GMsgk/src/GPUArraysCore.jl:225 [3] backend(x::Array{Float64, 0}) @ GPUArraysCore ~/.julia/packages/GPUArraysCore/GMsgk/src/GPUArraysCore.jl:226 [4] _copyto! @ ~/.julia/packages/GPUArrays/8Y80U/src/host/broadcast.jl:78 [inlined] [5] materialize! @ ~/.julia/packages/GPUArrays/8Y80U/src/host/broadcast.jl:38 [inlined] [6] materialize! @ ./broadcast.jl:911 [inlined] [7] macro expansion @ ~/.julia/packages/OMEinsum/pPXqJ/src/utils.jl:29 [inlined] [8] binary_einsum!(::OMEinsum.SimpleBinaryRule{(), (), ()}, x1::Array{Float64, 0}, x2::CuArray{Float64, 0, CUDA.DeviceMemory}, y::Array{Float64, 0}, sx::Bool, sy::Bool) @ OMEinsum ~/.julia/packages/OMEinsum/pPXqJ/src/binaryrules.jl:19 [9] einsum!(ixs::Vector{Vector{Int64}}, iy::Vector{Int64}, xs::Tuple{Any, Any}, y::Any, sx::Bool, sy::Bool, size_dict::Dict{Int64, Int64}) @ OMEinsum ~/.julia/packages/OMEinsum/pPXqJ/src/einsum.jl:115 [10] einsum! @ OMEinsum ~/.julia/packages/OMEinsum/pPXqJ/src/einsum.jl:38 [inlined] [11] einsum(code::DynamicEinCode{Int64}, xs::Tuple, size_dict::Dict{Int64, Int64}) @ OMEinsum ~/.julia/packages/OMEinsum/pPXqJ/src/einsum.jl:33 [12] einsum(code::DynamicEinCode{Int64}, xs::Tuple{RescaledArray{Float64, 0, Array{Float64, 0}}, RescaledArray{Float64, 0, CuArray{Float64, 0, CUDA.DeviceMemory}}}, size_dict::Dict{Int64, Int64}) @ TensorInference ~/.julia/dev/TensorInference/src/RescaledArray.jl:39 [13] einsum_grad(ixs::Vector{Vector{Int64}}, xs::Any, iy::Vector{Int64}, size_dict::Dict{Int64, Int64}, cdy::RescaledArray{Float64, 0, Array{Float64, 0}}, i::Int64) @ OMEinsum ~/.julia/packages/OMEinsum/pPXqJ/src/autodiff.jl:28 [14] #183 @ TensorInference ~/.julia/dev/TensorInference/src/mar.jl:82 [inlined] [15] ntuple @ Base ./ntuple.jl:19 [inlined] [16] einsum_backward_rule(eins::DynamicEinCode{Int64}, xs::Tuple{RescaledArray{Float64, 0, CuArray{Float64, 0, CUDA.DeviceMemory}}, RescaledArray{Float64, 0, CuArray{Float64, 0, CUDA.DeviceMemory}}}, y::RescaledArray{Float64, 0, CuArray{Float64, 0, CUDA.DeviceMemory}}, size_dict::Dict{Int64, Int64}, dy::RescaledArray{Float64, 0, Array{Float64, 0}}) @ TensorInference ~/.julia/dev/TensorInference/src/mar.jl:82 [17] generate_gradient_tree(code::DynamicNestedEinsum{Int64}, cache::TensorInference.CacheTree{Float64}, dy::RescaledArray{Float64, 0, Array{Float64, 0}}, size_dict::Dict{Int64, Int64}) @ TensorInference ~/.julia/dev/TensorInference/src/mar.jl:75 [18] generate_gradient_tree(se::SlicedEinsum{Int64, DynamicNestedEinsum{Int64}}, cache::TensorInference.CacheTree{Float64}, dy::RescaledArray{Float64, 0, Array{Float64, 0}}, size_dict::Dict{Int64, Int64}) @ TensorInference ~/.julia/dev/TensorInference/src/mar.jl:56 [19] gradient_tree(code::SlicedEinsum{Int64, DynamicNestedEinsum{Int64}}, xs::Vector{RescaledArray{Float64, N, AT} where {N, AT<:AbstractArray{Float64, N}}}) @ TensorInference ~/.julia/dev/TensorInference/src/mar.jl:94 [20] cost_and_gradient(code::SlicedEinsum{Int64, DynamicNestedEinsum{Int64}}, xs::Vector{RescaledArray{Float64, N, AT} where {N, AT<:AbstractArray{Float64, N}}}) @ TensorInference ~/.julia/dev/TensorInference/src/mar.jl:99 [21] marginals(tn::TensorNetworkModel{Int64, SlicedEinsum{Int64, DynamicNestedEinsum{Int64}}, Array{Float64}}; usecuda::Bool, rescale::Bool) @ TensorInference ~/.julia/dev/TensorInference/src/mar.jl:193 [22] macro expansion @ ./timing.jl:279 [inlined] [23] macro expansion @ ~/.julia/dev/TensorInference/test/cuda.jl:13 [inlined] [24] macro expansion @ ~/packages/julias/julia-1.10.0/share/julia/stdlib/v1.10/Test/src/Test.jl:1577 [inlined] [25] top-level scope @ ~/.julia/dev/TensorInference/test/cuda.jl:7 [26] include(fname::String) @ Base.MainInclude ./client.jl:489 [27] top-level scope @ ~/.julia/dev/TensorInference/test/runtests.jl:29 [28] include(fname::String) @ Base.MainInclude ./client.jl:489 [29] top-level scope @ none:6 [30] eval @ Core ./boot.jl:385 [inlined] [31] exec_options(opts::Base.JLOptions) @ Base ./client.jl:291 [32] _start() @ Base ./client.jl:552 Test Summary: | Error Total Time gradient-based tensor network solvers | 1 1 1m13.3s ERROR: LoadError: Some tests did not pass: 0 passed, 0 failed, 1 errored, 0 broken. in expression starting at /home/mroavi/.julia/dev/TensorInference/test/cuda.jl:6 in expression starting at /home/mroavi/.julia/dev/TensorInference/test/runtests.jl:28 ```

[mroavi]

and I'm getting the following error when trying to resolve the package versions:

(TensorInference) pkg> resolve
ERROR: empty intersection between GenericTensorNetworks@1.4.1 and project compatibility 2

[GiggleLiu]

Wield, have you used a different registry? and have you tried updating the packages? The CI passed.

[mroavi]

Wield, have you used a different registry? and have you tried updating the packages? The CI passed.

I have not used a different registry. I updated the packages but I'm still getting these error:

  Got exception outside of a @test
  This object is not a GPU array

Do the tests pass in your machine?

[GiggleLiu]

I just fixed the bug, it is about incorrectly converting a target array to a rescaled array on GPU. Thanks for checking the CUDA test!

[mroavi]

I just fixed the bug, it is about incorrectly converting a target array to a rescaled array on GPU. Thanks for checking the CUDA test!

All tests pass now! Awesome work.