ALM terminated unexpectedly

[Lamed-git]

Dear Dr. Tadano,
I use the latest version of ALAMODE (1.5.0), but I still encounter the ALM unexpected termination situation, I try to change FCSYM_BASIS to C, but the error still exists, the same supercell training the second order force constant does not report an error, but when I get the anharmonic interatomic force constants something goes wrong:

I checked the input file carefully to make sure that the content you mentioned earlier is OK：

1. ALM is able to correctly identify symmetries (No.164 space group), and 16 decimal places are used for both cell parameters and atomic sites.
2. The system itself has 108 atoms, 199 input files are used for training, the total free parameter is 29076, the memory used is 256 GB, there should be no memory overflow problem.
3. I adjusted the FCSYM_BASIS (L or C) and LMODEL (enet or adaptive-lasso), but it didn't help, the output still terminated at:
   

[ttadano]

Hi,

This issue is probably related to the numerical instability of the matrix reduction algorithm inside ALM, but please let me confirm this.

For the moment, please try setting TOL_CONST to larger or smaller than 1.0e-6 (default), which may change the behavior of the code.

[Lamed-git]

Hi,
I can't find TOL_CONST in the List of supported input variables, only TOLERANCE (default value: 1.0e-3) and CONV_TOL (default value: 1.0e-8) seem to be available at the moment. I tried to increase CONV_TOL to 1.0e-3 and ALM worked well. I'm not sure that's a reasonable adjustment.

[ttadano]

TOL_CONST is not included on the webpage, but it is available.

You can change it by writing, for example, TOL_CONST = 1.0e-8 in the &general field.

[Lamed-git]

I've solved this problem by change TOL_CONST = 1.0e-8.

[ttadano]

I'm delighted to hear that. Still, I will work on improving the stability of the algorithm.