Conversation
…l-valued C4v Heisenberg test
…tor` and add QR-CTMRG optimization tests
Codecov Report✅ All modified and coverable lines are covered by tests.
🚀 New features to boost your workflow:
|
|
After the build completes, the updated documentation will be available here |
Yue-Zhengyuan
left a comment
Yue-Zhengyuan
left a comment
There was a problem hiding this comment.
Minor comments on the example.
docs/src/examples/c4v_ctmrg/index.md
Outdated
| [ Info: LBFGS: iter 20, Δt 250.4 ms: f = -6.602282359103e-01, ‖∇f‖ = 1.2155e-03, α = 1.00e+00, m = 19, nfg = 1 | ||
| [ Info: LBFGS: iter 21, Δt 126.2 ms: f = -6.602299515427e-01, ‖∇f‖ = 1.0743e-03, α = 1.00e+00, m = 20, nfg = 1 | ||
| [ Info: LBFGS: iter 22, Δt 253.0 ms: f = -6.602310402232e-01, ‖∇f‖ = 4.7766e-04, α = 1.00e+00, m = 20, nfg = 1 | ||
| [ Info: LBFGS: converged after 23 iterations and time 2.06 s: f = -6.602310919804e-01, ‖∇f‖ = 5.4688e-05 |
There was a problem hiding this comment.
Honestly, QR-CTMRG performance is quite mundane in the example...
There was a problem hiding this comment.
Very true, so I didn't want to highlight this here. I think the reason is that at
Co-authored-by: Yue Zhengyuan <yuezy1997@icloud.com>
leburgel
left a comment
leburgel
left a comment
There was a problem hiding this comment.
The example itself looks great, I just have a few comments on the src changes. If you prefer merging the example first and then addressing these separately that would be fine by me. Having the example in is the most important, but then we should sort these out immediately after.
Co-authored-by: Lukas Devos <ldevos98@gmail.com>
|
I'll just rerun everything today. Probably doesn't hurt since some of the generated examples are quite old now. |
Could you just rerun the C4v ones here, and then do the other older generated examples in a different PR? Just so the PR title matches the actual git diff. |
4 participants
Here I add a small example where we demonstrate how to work with C4v-symmetric PEPS and environments and what to look out for when contracting and optimizing with C4v CTMRG. I also added a section on QR-CTMRG.
In order to make C4v optimization work with real numbers I needed to modify a small check in
fixedpoint(which wrongly stopped us from using real numbers with:fixedmode differentiation for C4v CTMRG). Also, I had to introduce a dummy truncation error and condition number to the returninfotuple inc4v_projector!to make QR-CTMRG work withfixedpoint. For both modifications I added tests. Maybe there is also a meaningful way to compute a truncation error/condition number for QR-CTMRG?