Recently I was running a 1D half-filling chain simulation with Coulomb interactions and I realized that even the ground state calculations could not be completed once the chain becomes fairly long (I've tried 80 for instance). On the other hand, this code runs perfectly fine for smaller system (I tried L=40 or smaller).
I ran the code using (in my last calculation)
python amn.py 80 1 3
The errors:
Execute_MPSMain MPS_TMP_LR/mps_long_range_MPS_L80_N40_t1.0_lambda1.0_nuc1_e2_N3Main.nml 0 running with 1 0 4KilledTraceback (most recent call last): File "amn.py", line 238, in <module> LongRange( L, lambdas, nfactor, Nst, MPS=MPSQ, PostProcess=PostProcessQ) File "amn.py", line 131, in LongRange mps.runMPS(MainFiles, RunDir=RunDir) File "/path/anaconda3/envs/openmps/lib/python3.7/site-packages/MPSPyLib/tools.py", line 398, in runMPS raise MPSFortLibError(ret_val)MPSPyLib.tools.MPSFortLibError: MPSFortLib quit with fatal return code 137!
Here, I suspect that error code 137 is just the SIGKILL code resulting from the earlier 'Killed' command, so I'm not entirely sure what is going on.
For your information, I'm running this code from a machine with i7-5500U CPU and 8GB ram. I will test it on a more powerful machine once I have the chance, if it has something to do with performance.
On a completely separate note, the previously mentioned L=40 calculation also wouldn't converge beyond n_excite_state >= 7 or more. Specifically, I'm seeing parameter 10 to DGEMM has illegal value.
Although I think this is easier to understand (Correct me if I'm wrong). As the more states are to be orthogonalized , the smaller the projected space becomes, so eventually the variational algorithm will fail become of precision issues. However it would still be nice if you have some suggestions so that I can push the code for even a couple more excited states.
Hi there,
Recently I was running a 1D half-filling chain simulation with Coulomb interactions and I realized that even the ground state calculations could not be completed once the chain becomes fairly long (I've tried 80 for instance). On the other hand, this code runs perfectly fine for smaller system (I tried L=40 or smaller).
I ran the code using (in my last calculation)
python amn.py 80 1 3The errors:
Here, I suspect that error code 137 is just the SIGKILL code resulting from the earlier 'Killed' command, so I'm not entirely sure what is going on.
For your information, I'm running this code from a machine with i7-5500U CPU and 8GB ram. I will test it on a more powerful machine once I have the chance, if it has something to do with performance.
On a completely separate note, the previously mentioned L=40 calculation also wouldn't converge beyond n_excite_state >= 7 or more. Specifically, I'm seeing parameter 10 to DGEMM has illegal value.
Although I think this is easier to understand (Correct me if I'm wrong). As the more states are to be orthogonalized , the smaller the projected space becomes, so eventually the variational algorithm will fail become of precision issues. However it would still be nice if you have some suggestions so that I can push the code for even a couple more excited states.