DMRG with adaptive number of sweeps

ITensor Julia Questions
1

Hi!
Is there a built-in way to run a dmrg simulation not with a fixed number of sweeps, but ending it when it converges up to user need?

As an example, I might need a precision of 10^-5, as soon as a new sweep is not improving the eigenvalue for more than this amount the simulation stops (or within the last 10 or something similar).

Thank you so much!

2

Is the question not clear? If so I can try to explain what I mean better

3

We have a feature built into our DMRG code called the observer system that lets you both monitor a calculation and stop it early based on any criterion you want. Here’s a walkthrough of how to use it:
<
https://itensor.github.io/ITensors.jl/stable/Observer.html>

Does this give you what you need? For your application you’d need to define a custom checkdone! function.

4

Thank you so much!