You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Summary
Add MultiSystems support while posting vasp jobs
Detailed Description
I am currently investigating the feasibility of incorporating multiple vacancy structures with different formulas into a single sys_configs tag. However, I encountered error message #306 due to inconsistent atom numbers in this mixed system setup. This behavior was expected, but I have identified a potential solution to address it.
To resolve the issue, I suggest modifying the post_fp_vasp function in dpgen/dpgen/generator/run.py:3914 as follows: replace all_sys = None with all_sys = dpdata.MultiSystems(type_map=jdata["type_map"]). This adjustment utilizes the MultiSystems functionality from the dpdata library, effectively preventing the single-labeled system complaining about formula difference.
While this change successfully resolves the problem, I have concerns regarding potential risks associated with introducing multiple systems in one fp task. Therefore, I would like to inquire about any potential harm or drawbacks in this approach. If there are no significant concerns, I kindly request your consideration in applying this code change, enabling the mixing of vacancy structures into one training system.
The text was updated successfully, but these errors were encountered:
Summary
Add MultiSystems support while posting vasp jobs
Detailed Description
I am currently investigating the feasibility of incorporating multiple vacancy structures with different formulas into a single sys_configs tag. However, I encountered error message #306 due to inconsistent atom numbers in this mixed system setup. This behavior was expected, but I have identified a potential solution to address it.
To resolve the issue, I suggest modifying the post_fp_vasp function in dpgen/dpgen/generator/run.py:3914 as follows: replace all_sys = None with all_sys = dpdata.MultiSystems(type_map=jdata["type_map"]). This adjustment utilizes the MultiSystems functionality from the dpdata library, effectively preventing the single-labeled system complaining about formula difference.
While this change successfully resolves the problem, I have concerns regarding potential risks associated with introducing multiple systems in one fp task. Therefore, I would like to inquire about any potential harm or drawbacks in this approach. If there are no significant concerns, I kindly request your consideration in applying this code change, enabling the mixing of vacancy structures into one training system.
The text was updated successfully, but these errors were encountered: