YaoTensorNetwork

Converting a Quantum Circuit in Yao@v0.6 to a tensor network.

To start, open a Julia REPL and type ] to enter pkg mode, install dependancies by

pkg> add Yao LuxurySparse BitBasis DelimitedFiles OMEinsum
pkg> dev YaoExtensions
pkg> dev git@github.com:QuantumBFS/YaoTensorNetwork.jl.git

If the second line does not work, please try clone and pkg> dev . at top level folder.

Learn by Example

julia> using Yao, YaoExtensions, YaoTensorNetwork

julia> c = dispatch!(variational_circuit(2, 1, [1=>2]), :random);

julia> eg = circuit2tn(c; initial_config=bit"00", final_config=bit"11")
EinGraph{Complex{Float64},Array{Complex{Float64},N} where N}
 T[1](2)
 T[2](2)
 T[1,3](2, 2)
 T[3,4](2, 2)
 T[2,5](2, 2)
 T[5,6](2, 2)
 T[4,7,8](2, 2, 2)
 T[6,9,8](2, 2, 2)
 T[7,10](2, 2)
 T[10,11](2, 2)
 T[9,12](2, 2)
 T[12,13](2, 2)
 T[11](2)
 T[13](2)


julia> dump_graph("_test", eg);

julia> eg2 = load_graph(eltype(eg), "_test");

julia> using OMEinsum

julia> res = contract(eg)
-0.005533928306495697 - 0.21124814706199962im

Here, circuit2tn convert a circuit to a "generalized tensor network" (or factor graph). In order to general reasonable structures, we suggestion using simplify_blocktypes(c) before dumping. dump_graph dumps this generated tensor network (the EinGraph instance) to three files, _test.labels.dat, _test.sizes.dat and _test.tensors.dat in plain text format. One can use load_graph to read these files. This package conditionally depends on OMEinsum, which is able to evaluate the tensor network directly utilizing @tensoropt defined in TensorOperations.jl. One can also load the data to python with the script in the example folder.

For more examples, see example folder.