J1Spin indexing issue #4612

shivupa · 2023-05-24T15:39:09Z

Proposed changes

This indexing in the spin dependent 1 body jastrow is wrong. I need to figure out a test that triggers this.

What type(s) of changes does this code introduce?

Bugfix

Does this introduce a breaking change?

No

What systems has this change been tested on?

AMD workstation

Checklist

Yes. This PR is up to date with current the current state of 'develop'
Yes. Code added or changed in the PR has been clang-formatted
No. This PR adds tests to cover any new code, or to catch a bug that is being fixed
No. Documentation has been added (if appropriate)

src/QMCWaveFunctions/Jastrow/J1Spin.h

quantumsteve · 2023-05-24T19:48:13Z

src/QMCWaveFunctions/Jastrow/J1Spin.h

+          for (int j = 0; j < NumTargetGroups; j++)
+            if (j == target_type && J1UniqueFunctors[igroup * NumTargetGroups + j] == nullptr)
+              J1UniqueFunctors[igroup * Nelec + j] = std::move(afunc);


Suggested change

for (int j = 0; j < NumTargetGroups; j++)

if (j == target_type && J1UniqueFunctors[igroup * NumTargetGroups + j] == nullptr)

J1UniqueFunctors[igroup * Nelec + j] = std::move(afunc);

if (J1UniqueFunctors[igroup * NumTargetGroups + target_type] == nullptr)

J1UniqueFunctors[igroup * Nelec + target_type] = std::move(afunc);

Why does the check use igroup * NumTargetGroups and the assignment use igroup * Nelec?

rcclay · 2023-05-24T20:04:04Z

How was this found?

shivupa · 2023-05-24T20:09:30Z

I hit this in a positron and Be cluster calculation. I am trying now to make a lightweight version of that system to use as a test

prckent · 2023-05-24T21:06:41Z

Important catch. Worth studying the J2 and J3 implementations for similar faults once this is resolved.

ye-luo

Can we have a unit test?

ye-luo · 2023-05-24T22:32:03Z

src/QMCWaveFunctions/Jastrow/J1Spin.h

-              J1UniqueFunctors[i * NumTargetGroups + j] == nullptr)
-            J1UniqueFunctors[igroup * Nelec + j] = std::move(afunc);
-        }
+      {


I don't see why loop i is even needed.
J1UniqueFunctors[source_type * NumTargetGroups + target_type]

If there were more than one combination of i and j then afunc would be moved multiple times.

Looks like that is happening on line 160 😟

Very true. target_type == -1 case seems broken. That needs to be sorted out when unifying J1 and J1Spin.

Yeah not sure what I was doing in the past but this was definitely wrong. target_type should only be -1 for spin independent J1 so the radial jastrow builder should never call this function with target_type == -1. I now raise a runtime error if we ever hit that case.

Also, yes, no loop is even needed.

shivupa · 2023-05-25T14:39:12Z

Still needs a test

shivupa · 2023-05-25T14:46:15Z

Important catch. Worth studying the J2 and J3 implementations for similar faults once this is resolved.

No issues in J2/J3

shivupa · 2023-06-01T20:25:40Z

Ok I added a test that fails on the current develop only for the cloned version of the Jastrow (which makes sense since I saw this error when performing a Jastrow optimization).


/home/shiv/Documents/qmcpack/src/QMCWaveFunctions/tests/test_J1Spin.cpp:228: FAILED:
  CHECK( cloned_log == LogComplexApprox(expected_log) )
with expansion:
  (-1.74258,0)
  ==
  LogComplexApprox( (-3.58983,0) )

/home/shiv/Documents/qmcpack/src/QMCWaveFunctions/tests/test_J1Spin.cpp:245: FAILED:
  CHECK( cloned_dlogpsi[i] == ValueApprox(expected_dlogpsi[i]) )
with expansion:
  0.0 == Approx( -2.544 )

/home/shiv/Documents/qmcpack/src/QMCWaveFunctions/tests/test_J1Spin.cpp:247: FAILED:
  CHECK( cloned_dhpsioverpsi[i] == ValueApprox(expected_dhpsioverpsi[i]) )
with expansion:
  -0.0 == Approx( -2.45001 )

/home/shiv/Documents/qmcpack/src/QMCWaveFunctions/tests/test_J1Spin.cpp:245: FAILED:
  CHECK( cloned_dlogpsi[i] == ValueApprox(expected_dlogpsi[i]) )
with expansion:
  0.0 == Approx( -4.70578 )

/home/shiv/Documents/qmcpack/src/QMCWaveFunctions/tests/test_J1Spin.cpp:247: FAILED:
  CHECK( cloned_dhpsioverpsi[i] == ValueApprox(expected_dhpsioverpsi[i]) )
with expansion:
  -0.0 == Approx( 0.0794429 )

/home/shiv/Documents/qmcpack/src/QMCWaveFunctions/tests/test_J1Spin.cpp:245: FAILED:
  CHECK( cloned_dlogpsi[i] == ValueApprox(expected_dlogpsi[i]) )
with expansion:
  0.0 == Approx( -0.055314 )

/home/shiv/Documents/qmcpack/src/QMCWaveFunctions/tests/test_J1Spin.cpp:247: FAILED:
  CHECK( cloned_dhpsioverpsi[i] == ValueApprox(expected_dhpsioverpsi[i]) )
with expansion:
  -0.0 == Approx( 0.0462761 )

/home/shiv/Documents/qmcpack/src/QMCWaveFunctions/tests/test_J1Spin.cpp:245: FAILED:
  CHECK( cloned_dlogpsi[i] == ValueApprox(expected_dlogpsi[i]) )
with expansion:
  0.0 == Approx( -0.770138 )

/home/shiv/Documents/qmcpack/src/QMCWaveFunctions/tests/test_J1Spin.cpp:247: FAILED:
  CHECK( cloned_dhpsioverpsi[i] == ValueApprox(expected_dhpsioverpsi[i]) )
with expansion:
  -0.0 == Approx( -0.330801 )

===============================================================================
test cases:   28 |  27 passed | 1 failed
assertions: 1005 | 996 passed | 9 failed

shivupa · 2023-06-01T21:54:48Z

Upload coverage step failing in CI. Seems unrelated

ye-luo · 2023-06-01T21:55:55Z

Test this please

ye-luo · 2023-06-01T21:56:35Z

Test this please

shivupa changed the title ~~Fix indexing issue~~ J1Spin indexing issue May 24, 2023

quantumsteve reviewed May 24, 2023

View reviewed changes

src/QMCWaveFunctions/Jastrow/J1Spin.h Outdated Show resolved Hide resolved

quantumsteve reviewed May 24, 2023

View reviewed changes

ye-luo requested changes May 24, 2023

View reviewed changes

shivupa added 7 commits June 1, 2023 14:03

Fix indexing issue

c1bf7ec

Format

eb9d689

cleaning

d0a390e

remove useless loop

933423d

Fix

e4defc9

Clean up addFunc

4de4279

Adding test

5b2d327

shivupa force-pushed the J1spin_indexing_fix branch from 02e14d9 to 5b2d327 Compare June 1, 2023 20:19

Merge branch 'develop' into J1spin_indexing_fix

0574de9

ye-luo approved these changes Jun 1, 2023

View reviewed changes

ye-luo enabled auto-merge June 1, 2023 21:57

ye-luo merged commit 58424b9 into QMCPACK:develop Jun 1, 2023

prckent mentioned this pull request Aug 18, 2023

Rc 3170 #4702

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

J1Spin indexing issue #4612

J1Spin indexing issue #4612

shivupa commented May 24, 2023

quantumsteve May 24, 2023 •

edited

Loading

quantumsteve May 24, 2023

rcclay commented May 24, 2023

shivupa commented May 24, 2023

prckent commented May 24, 2023

ye-luo left a comment

ye-luo May 24, 2023

quantumsteve May 25, 2023

ye-luo May 25, 2023

shivupa May 25, 2023

shivupa commented May 25, 2023

shivupa commented May 25, 2023

shivupa commented Jun 1, 2023

shivupa commented Jun 1, 2023 •

edited

Loading

ye-luo commented Jun 1, 2023

ye-luo commented Jun 1, 2023

J1Spin indexing issue #4612

J1Spin indexing issue #4612

Conversation

shivupa commented May 24, 2023

Proposed changes

What type(s) of changes does this code introduce?

Does this introduce a breaking change?

What systems has this change been tested on?

Checklist

quantumsteve May 24, 2023 • edited Loading

Choose a reason for hiding this comment

quantumsteve May 24, 2023

Choose a reason for hiding this comment

rcclay commented May 24, 2023

shivupa commented May 24, 2023

prckent commented May 24, 2023

ye-luo left a comment

Choose a reason for hiding this comment

ye-luo May 24, 2023

Choose a reason for hiding this comment

quantumsteve May 25, 2023

Choose a reason for hiding this comment

ye-luo May 25, 2023

Choose a reason for hiding this comment

shivupa May 25, 2023

Choose a reason for hiding this comment

shivupa commented May 25, 2023

shivupa commented May 25, 2023

shivupa commented Jun 1, 2023

shivupa commented Jun 1, 2023 • edited Loading

ye-luo commented Jun 1, 2023

ye-luo commented Jun 1, 2023

quantumsteve May 24, 2023 •

edited

Loading

shivupa commented Jun 1, 2023 •

edited

Loading