This is not actually a question, just something that I came across and thought might be interesting to the users.
I have a system of four species (F, Li, Be and Cr) where Cr is an impurity and there is only one of it in the unit cell. My unit cell size is about 11 angstrom and I'm using cutoff radius of 10 angstrom for training (so the atoms won't see their own periodic images).
Since there is only 1 atom of Cr in the system, the the parameters for Cr-Cr end up being zero after the training due to no information about their interaction, which is fine. The problem is when I'm trying to generate an active set I'm getting the following error:
###########################
Traceback (most recent call last):
File "/home/sattarian/.local/bin/pace_activeset", line 169, in
A_active_set_dict = compute_active_set(A0_proj_dict, tol=gamma_tolerance, max_iters=maxvol_iters, verbose=True)
File "/home/sattarian/.local/lib/python3.8/site-packages/pyace/activelearning.py", line 285, in compute_active_set
selectedrows, = maxvol(cur_A0, tol=tol, max_iters=max_iters)
File "/home/sattarian/.local/lib/python3.8/site-packages/maxvolpy/maxvol.py", line 285, in maxvol
return maxvol_func(A, tol=tol, max_iters=max_iters,
File "maxvolpy/_maxvol.pyx", line 114, in maxvolpy._maxvol.c_maxvol
File "maxvolpy/_maxvol.pyx", line 103, in maxvolpy._maxvol.c_maxvol
File "maxvolpy/_maxvol.pyx", line 376, in maxvolpy._maxvol.dmaxvol
ValueError: Input matrix must not be singular
###########################
After playing with the values in the output_potential.yaml, I learned that I need to increase the cutoff radius for Cr-Cr interactions to something higher than 11 (15 for example) temporarily when doing the active set generation to overcome this issue.
you may want to add a warning or something to check for such scenario before hand so the user won't be confused when facing this error
This is not actually a question, just something that I came across and thought might be interesting to the users.
I have a system of four species (F, Li, Be and Cr) where Cr is an impurity and there is only one of it in the unit cell. My unit cell size is about 11 angstrom and I'm using cutoff radius of 10 angstrom for training (so the atoms won't see their own periodic images). Since there is only 1 atom of Cr in the system, the the parameters for Cr-Cr end up being zero after the training due to no information about their interaction, which is fine. The problem is when I'm trying to generate an active set I'm getting the following error:
########################### Traceback (most recent call last): File "/home/sattarian/.local/bin/pace_activeset", line 169, in
A_active_set_dict = compute_active_set(A0_proj_dict, tol=gamma_tolerance, max_iters=maxvol_iters, verbose=True)
File "/home/sattarian/.local/lib/python3.8/site-packages/pyace/activelearning.py", line 285, in compute_active_set
selectedrows, = maxvol(cur_A0, tol=tol, max_iters=max_iters)
File "/home/sattarian/.local/lib/python3.8/site-packages/maxvolpy/maxvol.py", line 285, in maxvol
return maxvol_func(A, tol=tol, max_iters=max_iters,
File "maxvolpy/_maxvol.pyx", line 114, in maxvolpy._maxvol.c_maxvol
File "maxvolpy/_maxvol.pyx", line 103, in maxvolpy._maxvol.c_maxvol
File "maxvolpy/_maxvol.pyx", line 376, in maxvolpy._maxvol.dmaxvol
ValueError: Input matrix must not be singular
###########################
After playing with the values in the output_potential.yaml, I learned that I need to increase the cutoff radius for Cr-Cr interactions to something higher than 11 (15 for example) temporarily when doing the active set generation to overcome this issue.
you may want to add a warning or something to check for such scenario before hand so the user won't be confused when facing this error
Thanks, Siamak