-
Notifications
You must be signed in to change notification settings - Fork 139
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Integration test needed for the ACForce estimator #2238
Comments
Sounds like a good target for addtional unit/deterministic/integration tests. |
Ack! A simple view of this would be that we need a simple integration test and preferably a deterministic integration test that uses ACForce. However, better would be a unit or small scale integration test that can catch this. |
#2232 will be done today and should fix this |
@tiihonej Please modify an existing test, e.g. LiH dimer. This will avoid introducing new files for the wavefunctions and potentials. Of course, this assumes that LiH triggers the test. I suspect that parsing the force to check the numerical results will require writing some new python code in the test system, but for now that is not essential: the key test is whether the code crashes or not. |
@prckent Modifying existing file should be alright and, and LiH already covers a great deal of force-related features (the only exception probably being the derivatives of some higher-order WF-components). Testing against the values of the Pulay terms should be straightforward and catch all there is to catch regarding the actual performance. |
@tiihonej Do you mean you get the same resize error? I'm running your example with this command line OMP_NUM_THREADS=1 mpirun --mca btl tcp,self -np 16 -map-by numa:PE=1 ../bin/qmcpack vmc.force.xml |
@PDoakORNL Nevermind that, there was something else wrong with my build. I did a fresh checkout and everything works. I'll push an integration test and then close this issue. |
There are a few issues with adding an integration test for the current ACForce estimator. First, none of the needed observables (ACForce_{hf,pulay,wfgrad, Ewfgrad}) are implemented in the checking macros or nexus. Adding them is perhaps a bit tedious, considering that the final observable format is perhaps subject to discussion (ie, is scalar.dat a proper place for tensorial quantities? Should QMCPACK sum up the pulay corrections internally and only/also output the final force?). Second, LiH_dimer_pp is a good and simple platform for the test, but the wavefunction is too good! At equilibrium we get essentially zero forces and corrections, which is not so good for testing. I think a test like this should only be added once we are done with streamlining the code, including nexus support. |
QMCPACK runs with the ACForce estimator crash at launch with the latest code (since around Jan 16). The error message is something like the following:
I take suggestions to better pinpoint where this is coming from. Sample inputs are attached:
vmc_force_jan2020.tar.gz
The text was updated successfully, but these errors were encountered: