When I change the Julia ITensor to the latest version, an error arises for my mixed site model, which works for previous versions. And I have not updated julia ITensor
for several months.
The minimal code I have tested is as follows:
using ITensors
function mixedSites(N::Int)
sites = Vector{Index}(undef,N)
for n=1:N
if isodd(n)
sites[n] = siteind("Electron"; addtags="n=$n", conserve_qns=true)
else
sites[n] = siteind("Boson"; addtags ="n=$n", dim = 4, conserve_qns=false)
end
end
return sites
end
let
N = 40
sites = mixedSites(N)
state = Vector{String}(undef, N)
for xy = 1 : N
if xy % 2 == 1 #elecreon site
state[xy] = trunc(xy/2) % 2 == 0 ? "Up" : "Dn"
else #Boson site
state[xy] = "1"
end
end
psi0 = randomMPS(sites, state; linkdims = 20)
end
There is an error " LoadError: MethodError: no method matching sort(::Tuple{Int64, Int64})", which arises from the randomMPS() function.
It should be pointed out that I already set the space in qudits.jl to return QN(qnname_number, 0) =>dim even if I set conserve_qns = false for “Boson” to use the QN of “Electron”. I think that such a modification will not alert the result.
You’re right about needing to also modify the space function in qudits.jl or to use a custom one. But I think that could be the source of the error. Could you please also share how you modified it?
I found that the following modification of your code (without needing to change the space function) worked correctly:
# Changing the following line:
#sites[n] = siteind("Boson"; addtags ="n=$n", dim = 4, conserve_qns=false)
# Into this line:
sites[n] = Index(QN()=>4; tags="Boson,Site,n=$n")
is my original code, referred from a previous discussion section. When I changed the Boson site as above， the problem still unsolved.
The following is my modification.
function ITensors.space(
::SiteType"Qudit";
dim=2,
conserve_qns=false,
conserve_number=conserve_qns,
qnname_number="Number",
)
if conserve_number
return [QN(qnname_number, n - 1) => 1 for n in 1:dim]
else
return [QN(qnname_number, 0) => dim]
end
return dim
end
I see what you’re asking & it’s a good question. The difference I see is that you are writing
whereas I just used QN()=>dim. Could you please try that version with an empty argument to QN to see if that works?
I’m a bit confused myself because I would think those act rather similarly but maybe there’s a subtle difference that is mattering here. Also QN() better expresses the idea of “no quantum number”.
It seems that the problem is due to the Julia package itself. When I change julia-1.10.2 to julia-1.8.5. The error disappears but with a warning: “MPS center bond dimension is less than requested (you requested 12, but in practice, it is 1. This is likely due to technicalities of truncating quantum number sectors.”
Ideally users should never have to downgrade their Julia version to make ITensor code work, if that’s the case it should be considered a bug that we need to investigate.
@zhouzsh I would like to understand better why you had to downgrade your version of Julia to get things working. I can’t reproduce that issue you are seeing, here is what I am running:
using ITensors
function ITensors.space(
::SiteType"Qudit";
dim=2,
conserve_qns=false,
conserve_number=conserve_qns,
qnname_number="Number",
)
if conserve_number
return [QN(qnname_number, n - 1) => 1 for n in 1:dim]
else
return [QN() => dim]
end
return dim
end
function mixed_sites(N::Int)
sites = Vector{Index}(undef,N)
for n=1:N
if isodd(n)
sites[n] = siteind("Electron"; addtags="n=$n", conserve_qns=true)
else
sites[n] = siteind("Boson"; addtags ="n=$n", dim = 4, conserve_qns=false)
end
end
return sites
end
let
N = 40
sites = mixed_sites(N)
state = Vector{String}(undef, N)
for xy = 1 : N
if xy % 2 == 1 #elecreon site
state[xy] = trunc(xy/2) % 2 == 0 ? "Up" : "Dn"
else #Boson site
state[xy] = "1"
end
end
psi0 = randomMPS(sites, state; linkdims = 20)
end
which outputs an MPS without any errors, though with the warning:
Warning: MPS center bond dimension is less than requested (you requested 20, but in practice it is 1. This is likely due to technicalities of truncating quantum number sectors.
The versions I’m using are:
julia> using Pkg; Pkg.status()
Status `~/Simons Foundation Dropbox/Matthew Fishman/Documents/workdir/ITensors.jl/discourse/1619_sort_bug/Project.toml`
[9136182c] ITensors v0.3.66
julia> versioninfo()
Julia Version 1.10.2
Commit bd47eca2c8a (2024-03-01 10:14 UTC)
Build Info:
Official https://julialang.org/ release
Platform Info:
OS: macOS (arm64-apple-darwin22.4.0)
CPU: 10 × Apple M1 Max
WORD_SIZE: 64
LIBM: libopenlibm
LLVM: libLLVM-15.0.7 (ORCJIT, apple-m1)
Threads: 1 default, 0 interactive, 1 GC (on 8 virtual cores)
Does that code reproduce the issue you are seeing in Julia 1.10.2? If so, could you share the output of using Pkg; Pkg.status()?
Thanks for testing my code. Now I produce the same results as yours even if I upgrade my Julia package to 1.10.2. In fact, I have never modified the test code since the last time. Maybe it is due to the environment configuration issue after upgrading Julia and ITensor packages at the same time.