I have tried setting outputlevel=2
in tdvp
, but it produces the same output as outputlevel=1
. This option works correctly in dmrg
.
My versioninfo:
(@v1.10) pkg> status "ITensors"
Status `C:\Users\swan\.julia\environments\v1.10\Project.toml`
[9136182c] ITensors v0.6.11
(@v1.10) pkg> status "ITensorMPS"
Status `C:\Users\swan\.julia\environments\v1.10\Project.toml`
[0d1a4710] ITensorMPS v0.2.4
(@v1.10) pkg> status "ITensorTDVP"
Status `C:\Users\swan\.julia\environments\v1.10\Project.toml`
[25707e16] ITensorTDVP v0.4.4
julia> VERSION
v"1.10.4"
Minimal example is from ITensorMPS.jl/examples/01_tdvp.jl at main · ITensor/ITensorMPS.jl (github.com).
The reason I need the information at each site is that I’m calculating a rather challenging (long-range) model which is extremely sensitive to the initial state. I want to apply tdvp
to perform the imaginary time evolution as a warm-up and then proceed with DMRG. However, the tdvp
is extremely slow and even fails when combining QNs due to attempting to factorize/exponentiate a matrix containing NaNs/Infs
. But whatever, these questions may come as some follow-up posts.