ML_FF run stops after entering main loop

[akretschmer]

Hello,

I have trained a force field for a solid/liquid interface and want to do a production run. The initialization runs as normal, and when the main loop starts, the job exits. I do not see a relevant error message anywhere. VASP produces the timing information at the end of OUTCAR without really having done anything. The CONTCAR and OSZICAR files are empty.

What puzzles me is that the same calculation has run fine before with slightly lower values for NSW, ML_IERR and ML_OUTBLOCK. I have started this longer run a few times and always get the same behavior.

ML-FF-run.zip

[akretschmer]

I decreased ML_IERR and ML_OUTBLOCK from 30000 to 3000, and now it runs fine. So it seems that either of the two tags cannot use too large values.

[ferenc_karsai]

Thank you for posting your problem. I will have a look into it. If it's simple I will post the fix here otherwise you have to wait for the fix in the next update.