Update document #144

liwt31 · 2022-10-26T11:02:35Z

Fix the Heisenberg tutorial bug
Add time evolution page. Change the default time evolution method #143
Add the FAQ page

update @ 2023.4.8 @jjren

remove the m_max argument in _update_mps in mp.py.
First, because it is duplicated with mps.compress_config.
The more important reason is that if we want to adaptively control the bond dimension by truncation threshold in the two-site tdvp-ps or ground state algorithm, the former implementation can not achieve that.
(What I haven't modified is OFS algorithm because one loss function depends on m_max, so I added an assertion to make sure the bond dimension is fixed.)
add scipy ivp solver company to Krylov solver in the tdvp related evolution algorithm.
This will make the evolution algorithm suitable for non-Hermitian operators.
add quadrature in SineDVR basis by using the sympy expression to support more operators.
This feature is only experimental, not fully tested.
bump up dependencies (numpy, scipy) in requirement.txt

liwt31 · 2022-10-26T11:06:49Z

I may work on the tutorial module while you finish the time evolution part. @jjren

codecov · 2022-10-28T12:59:47Z

Codecov Report

Patch coverage: 79.38% and project coverage change: -0.53 ⚠️

Comparison is base (87dc576) 85.27% compared to head (4a03f1d) 84.74%.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #144      +/-   ##
==========================================
- Coverage   85.27%   84.74%   -0.53%     
==========================================
  Files         105      105              
  Lines        9995    10140     +145     
==========================================
+ Hits         8523     8593      +70     
- Misses       1472     1547      +75

Impacted Files	Coverage Δ
renormalizer/mps/mp.py	`84.75% <42.85%> (-3.81%)`	⬇️
renormalizer/spectra/base.py	`95.65% <75.00%> (-4.35%)`	⬇️
renormalizer/model/basis.py	`88.77% <82.97%> (-0.55%)`	⬇️
renormalizer/mps/gs.py	`95.81% <86.66%> (-1.28%)`	⬇️
renormalizer/mps/mps.py	`90.52% <92.85%> (-0.03%)`	⬇️
renormalizer/cv/spectra_cv.py	`88.88% <100.00%> (+0.09%)`	⬆️
renormalizer/cv/tests/test_H_chain.py	`100.00% <100.00%> (ø)`
renormalizer/cv/zerot.py	`98.14% <100.00%> (ø)`
renormalizer/lib/tests/test_krylov.py	`100.00% <100.00%> (ø)`
renormalizer/model/tests/test_basis.py	`100.00% <100.00%> (ø)`
... and 7 more

... and 2 files with indirect coverage changes

☔ View full report in Codecov by Sentry.
📢 Do you have feedback about the report comment? Let us know in this issue.

jjren · 2023-04-08T03:24:20Z

My another question is what the difference between the three CI tests.
From my understanding, ci / test (pull request) is the PR branch + master, ci / test (push) is the PR branch only. Then what is ci/circleci ?

jjren · 2023-04-08T04:05:39Z

Hi Weitang @liwt31 , at the beginning of this PR, you mentioned the Heisenberg tutorial bug, but I didn't find any changes to the Heisenberg tutorial. What I found is that the jupyter notebook of Heisenberg tutorial is supposed to move into doc/source/tutorial, but it doesn't. Did I do something wrong during the force push ... or you haven't move the jupyter notebook of Heisenberg into doc/source/tutorial?

I found that in the successful ci/circleci run, it showed that
Running Sphinx v5.3.0
copying ../../example/1D-Heisenberg.ipynb
making output directory... done
How did it happen? I can't find a rule anywhere in the package to do this.

Oh, finally, I found it in /home/jjren/Code/Renormalizer/doc/source/conf.py !
The remaining question is that

the bug in Heisenberg model tutorial
pandoc seems to have not been installed successfully in the two ci/test.

liwt31 · 2023-04-10T08:07:25Z

CI/CircleCI is based on the PR branch only. We are in the middle of switching from circleci to GitHub actions.
I can't really remember the bug exactly. Probably related to documentation generation and not the content in the notebook
CircleCI passes because in this branch the installation of pandoc is added to .circleci/config.yaml. The same lines should be added to .github/workflow/ci.yaml

liwt31 · 2023-04-10T08:26:27Z

renormalizer/__init__.py

@@ -6,7 +6,7 @@
 import warnings


-reno_num_threads = os.environ.get("RENO_NUM_THREADS")
+reno_num_threads = os.environ.get("RENO_NUM_THREADS", 1)


The default behavior is changed on purpose in #132 . I strongly oppose setting the default number of threads when importing renormalizer. The reason is that this is a self-centered feature and is bad for the applicability of renormalizer.

When renormalizer is imported as a regular package for another project (such as TenCirChem), setting the number of threads by default can cause unexpected behavior. For example, TenCirChem itself may have set the "MKL_NUM_THREADS" to a certain value for its purpose, and an experienced user can also set "MKL_NUM_THREADS" directly without referring to the documentation or source code of renormalizer. But the value will be overwritten silently by renormalizer when importing the package and there's not even a good way to prevent this behavior.

Also, the best practice for production-level calculation is to set the number of threads to 1, 2 or 4 based on the details of the calculation. Setting the default number of threads to 1 is more like a development shortcut. In fact, in my development environment, I hardcoded the number of threads to be 1 in __init__.py but I never commit this change.

This is also for consistency with popular computational software such as pyscf (pyscf/pyscf#540)

Got it. I will recover it.

renormalizer/cv/spectra_cv.py

liwt31 · 2023-04-10T08:40:44Z

renormalizer/model/basis.py

        elif set(op_symbol.split(" ")) == set("x"):
            moment = len(op_symbol.split(" "))
+            self._recursion_flag = 1


Is it better to use self._recursion_flag += 1 just as BasisSHO for deep recursion?

ok. I will do it.

liwt31 · 2023-04-10T08:49:06Z

renormalizer/mps/gs.py

-        logger.debug(f"mmax, percent: {mmax}, {percent}")
+
+        if isinstance(compress_config, CompressConfig):
+            mps.compress_config = compress_config


What about the original mps.compress_config? Should we record it and then set it back after the optimization?

We control the bond dimension in gs.py with mmax and the original mps.compress_config is not used. I guess it is the default one. So is it necessary to set it back?

The user may have set this value for future time evolution. pseudo-code:

mps.compress_config = ... mps.optimize_config = ... mps = optimize_mps(mps, mpo, procedure) mps = excitation_mpo @ mps mps = mps.evolve_with_rk(tau)

liwt31 · 2023-04-10T09:02:45Z

renormalizer/mps/mps.py

@@ -1190,9 +1199,21 @@ def func1(y):
                                    coef, ovlp_inv1=S_L_inv_list[imps+1],
                                    ovlp_inv0=S_L_inv_list[imps], ovlp0=S_L_list[imps])
                            return func(0, y)
+


Is it better to wrap this routine as a function?

Did you mean func1 is not necessary?

I mean L1203 to L1215, which has a lot of duplication in this commit

It is already a one-line function, although a bit longer.
I didn't think of a better way to make it elegant.

liwt31 · 2023-04-10T09:23:35Z

In general looks good! Only need to resolve the problem of the default thread number.

… non-hermitian H

jjren · 2023-07-03T15:52:45Z

Finally. 🐢

liwt31 · 2023-07-04T02:01:36Z

great stuff, thanks! Just a reminder: the document still needs a lot of work.

jjren force-pushed the doc branch from 297a962 to 981f4f7 Compare April 8, 2023 02:37

This comment was marked as outdated.

Sign in to view

liwt31 commented Apr 10, 2023

View reviewed changes

renormalizer/cv/spectra_cv.py Show resolved Hide resolved

liwt31 commented Apr 10, 2023

View reviewed changes

/bin/eash and others added 5 commits July 3, 2023 22:28

update doc

74ad9a8

add model tutorial

218073b

several changes: 1. _update_mps 2. SineDVRbasis 3. RK45 TD solver for…

d560333

… non-hermitian H

add comment

b214f66

update

4a03f1d

jjren force-pushed the doc branch from 8c6f992 to 981f4f7 Compare July 3, 2023 15:48

jjren force-pushed the doc branch from 981f4f7 to 4a03f1d Compare July 3, 2023 16:02

liwt31 merged commit 977eb2a into master Jul 4, 2023

liwt31 deleted the doc branch July 4, 2023 02:02

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Update document #144

Update document #144

liwt31 commented Oct 26, 2022 •

edited by jjren

Loading

liwt31 commented Oct 26, 2022

codecov bot commented Oct 28, 2022 •

edited

Loading

This comment was marked as outdated.

jjren commented Apr 8, 2023

This comment was marked as outdated.

jjren commented Apr 8, 2023 •

edited

Loading

liwt31 commented Apr 10, 2023

liwt31 Apr 10, 2023

liwt31 Apr 10, 2023

jjren Apr 10, 2023

jjren Jul 3, 2023 •

edited

Loading

liwt31 Apr 10, 2023

jjren Apr 10, 2023

jjren Jul 3, 2023

liwt31 Apr 10, 2023 •

edited

Loading

jjren Apr 10, 2023

liwt31 Apr 10, 2023

jjren Apr 10, 2023

jjren Jul 3, 2023

liwt31 Apr 10, 2023

jjren Apr 10, 2023

liwt31 Apr 10, 2023

jjren Jul 3, 2023

liwt31 commented Apr 10, 2023

jjren commented Jul 3, 2023

liwt31 commented Jul 4, 2023 •

edited

Loading

Update document #144

Update document #144

Conversation

liwt31 commented Oct 26, 2022 • edited by jjren Loading

liwt31 commented Oct 26, 2022

codecov bot commented Oct 28, 2022 • edited Loading

Codecov Report

This comment was marked as outdated.

jjren commented Apr 8, 2023

This comment was marked as outdated.

jjren commented Apr 8, 2023 • edited Loading

liwt31 commented Apr 10, 2023

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

jjren Jul 3, 2023 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

liwt31 Apr 10, 2023 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

liwt31 commented Apr 10, 2023

jjren commented Jul 3, 2023

liwt31 commented Jul 4, 2023 • edited Loading

liwt31 commented Oct 26, 2022 •

edited by jjren

Loading

codecov bot commented Oct 28, 2022 •

edited

Loading

jjren commented Apr 8, 2023 •

edited

Loading

jjren Jul 3, 2023 •

edited

Loading

liwt31 Apr 10, 2023 •

edited

Loading

liwt31 commented Jul 4, 2023 •

edited

Loading