Menu
▾
▴
#41 Excited states issue
Hello all,
I am using OSMPS v3 to try and calculate excited states of the Hubbard model. My code works fine for the ground state only, but if I add
'n_excited_states': 1,
'eMPSConvergenceParameters' : myConv,
'eMPSObservables' : myObservables,
to the set of parameters I get
Program received signal SIGSEGV: Segmentation fault - invalid memory reference.
Backtrace for this error:
0 0x7FEED8F0AE08
1 0x7FEED8F09F90
2 0x7FEED863B4AF
3 0x40377C in __basicops_MOD_descending_hsort
4 0x5031DB in __decompositionops_MOD_split_qtensor
5 0x6F8506 in __variationalops_MOD_find_excited_two_step_qmps_qmpo.constprop.2 at VariationalOps.f90:?
6 0x722994 in __variationalops_MOD_find_excited_two_qmps_qmpo
7 0x7EB85D in __pyinterface_MOD_static_simulation_mps_qmpo.constprop.12 at PyInterface.f90:?
8 0x80A338 in __pyinterface_MOD_simulation_qmpo.constprop.8 at PyInterface.f90:?
9 0x81F4E5 in __pyinterface_MOD_runmps
Segmentation fault (core dumped)
Traceback (most recent call last):
File "FermiHubbard_Exc_test.py", line 100, in <module>
mps.runMPS(MainFiles, RunDir='')
File "/usr/local/lib/python3.5/dist-packages/MPSPyLib/tools.py", line 394, in runMPS
raise MPSFortLibError(ret_val)
MPSPyLib.tools.MPSFortLibError: MPSFortLib quit with fatal return code 139!</module>
Weirdly, I recall that I could manage to get excited states with pretty much the same procedure in some previous version of OSMPS.
Can anyone help?
Kind regards,
Rafael
I've changed ticket ownership to myself.
Thanks for posting your issue on the forums. Would you mind sharing the python script you used to produce this pathology?
Additionally, if you have information on the last version of OpenMPS you used successfully, that would be illuminating. Even the last time (rough estimate of date) you did this and it worked is useful.
My guess is that an internal reference we relied on for this has changed in a more recent version. If that's the case, the fix should be straight forward.
Hi Matthew,
Thanks for the reply. The script is attached. Sometimes I am able to get results by reducing the size of the system L, but I don't recall that being th case previously.
The script is based on a version made available by D. Jaschke here (I am guessing for v3.0)
https://sourceforge.net/p/openmps/discussion/tech/thread/2bc53b16/
Best wishes,
Rafael
I'll have a look. Thanks!
I'll write back on Friday with an update
Hi Matthew,
Any update on this topic?
Best wishes,
Rafael
Hi guys,
Any chance of still getting some feedback here?
Best regards,
Rafael
Thank you for you're tremendous patience, Rafael!
I believe I've found the source of the bug and am writing a fix right now. My hope is that this fix addressees the core issue while I draft code that satisfies the larger design goals for the project.
The prototype, I hope to have implemented this week, perhaps Monday. I was planning on sending you a patch file that you can apply on your download.
Hi Matthew,
No problem, thanks for the reply. It would be surely nice to see this working again. Also, if possible I would like some more detailed instructions on the mac installation (see the other ticket), since I have not been able to solve that issue as well.
Best wishes,
Rafael
Hi Rafael,
I took a quick look at the issue, could reproduce it with some easier
parameters (L=40, chi=15, 5 sweeps, and 1 excited state). The new master on
SourceForge fixes the error for these parameters and hopefully yours. Can
you check if this solves your problem? If not, some easier parameters to
reproduce the bug would be convenient. Sorry that I was not looking into
the ticket before.
Best regards,
Daniel
On Mon, Apr 6, 2020 at 11:53 AM Rafael Barfknecht rafael-eb@users.sourceforge.net wrote:
Related
Tickets: #41
Hello Daniel,
Thank you for your work! I am not able to verify that it works right now, but as soon as possible I will provide some feedback on this.
Best regards,
Rafael