Fix TensorNetwork serialization

[jofrevalles]

This PR fixes the serialization problem of TensorNetworks. Since this function was not tested until now, I don't know how long it has not been working. This PR also adds a testset to cover this function.

In the following code snippets, I show the serialization problem with an MPS, but the same problem holds for any TensorNetwork.

Current problem

The deserialize function does not work:

julia> using Tenet; using Serialization

julia> mps = rand(Chain, Open, State; n=10, χ=10)
MPS (inputs=0, outputs=10)

julia> buffer = IOBuffer()
IOBuffer(data=UInt8[...], readable=true, writable=true, seekable=true, append=false, size=0, maxsize=Inf, ptr=1, mark=-1)

julia> serialize(buffer, mps)

julia> seekstart(buffer)
IOBuffer(data=UInt8[...], readable=true, writable=true, seekable=true, append=false, size=7794, maxsize=Inf, ptr=1, mark=-1)

julia> content = read(buffer)
7794-element Vector{UInt8}:
...

julia> read_buffer  = IOBuffer(content)
IOBuffer(data=UInt8[...], readable=true, writable=false, seekable=true, append=false, size=7794, maxsize=Inf, ptr=1, mark=-1)

julia> mps2 = deserialize(read_buffer)
ERROR: MethodError: no method matching deserialize(::Serializer{IOBuffer}, ::Vector{Tensor})

Closest candidates are:
  deserialize(::AbstractSerializer, ::Type{T}) where T<:Base.GenericCondition
   @ Serialization ~/.julia/juliaup/julia-1.10.5+0.x64.linux.gnu/share/julia/stdlib/v1.10/Serialization/src/Serialization.jl:1592
  deserialize(::AbstractSerializer, ::Type{T}) where T<:Distributed.WorkerPool
   @ Distributed ~/.julia/juliaup/julia-1.10.5+0.x64.linux.gnu/share/julia/stdlib/v1.10/Distributed/src/workerpool.jl:69
  deserialize(::AbstractSerializer, ::Type{T}) where T<:Base.AbstractLock
   @ Serialization ~/.julia/juliaup/julia-1.10.5+0.x64.linux.gnu/share/julia/stdlib/v1.10/Serialization/src/Serialization.jl:1580
  ...

Stacktrace:
...

With this PR

This PR solves the problem:

julia> using Tenet; using Serialization

julia> mps = rand(Chain, Open, State; n=10, χ=10)
MPS (inputs=0, outputs=10)

julia> buffer = IOBuffer()
IOBuffer(data=UInt8[...], readable=true, writable=true, seekable=true, append=false, size=0, maxsize=Inf, ptr=1, mark=-1)

julia> serialize(buffer, mps)

julia> seekstart(buffer)
IOBuffer(data=UInt8[...], readable=true, writable=true, seekable=true, append=false, size=7836, maxsize=Inf, ptr=1, mark=-1)

julia> content = read(buffer)
7836-element Vector{UInt8}:
...

julia> read_buffer  = IOBuffer(content)
IOBuffer(data=UInt8[...], readable=true, writable=false, seekable=true, append=false, size=7836, maxsize=Inf, ptr=1, mark=-1)

julia> mps2 = deserialize(read_buffer)
MPS (inputs=0, outputs=10)

julia> mps == mps2
true

[mofeing]

mmm I'm trying to understand this line. Is this needed in order to tell Julia that the object to deserialize is a TensorNetwork right?

[jofrevalles]

Yes exactly, this is what I have seen in similar posts (for example).

[mofeing]

Thanks!