Inconsistent atom numbers (formula) in the starting configurations of Exploration process

[DingChangjie]

Dear developers : when working on an alloy system, I tried to put some defect configurations (including vacancies and self-interstitial-atoms) into the starting configurations of Exploration. It all worked well until the last step, i.e. the "post_train". A Runtime Error occurred as follows :
INFO:dpgen:failed tasks: 0 in 5 0.00 % Traceback (most recent call last): File "/public/home/hfcas_user39/.local/bin/dpgen", line 8, in <module> sys.exit(main()) File "/public/home/hfcas_user39/.local/lib/python3.9/site-packages/dpgen/main.py", line 175, in main args.func(args) File "/public/home/hfcas_user39/.local/lib/python3.9/site-packages/dpgen/generator/run.py", line 2410, in gen_run run_iter (args.PARAM, args.MACHINE) File "/public/home/hfcas_user39/.local/lib/python3.9/site-packages/dpgen/generator/run.py", line 2399, in run_iter post_fp (ii, jdata) File "/public/home/hfcas_user39/.local/lib/python3.9/site-packages/dpgen/generator/run.py", line 2267, in post_fp post_fp_vasp(iter_index, jdata) File "/public/home/hfcas_user39/.local/lib/python3.9/site-packages/dpgen/generator/run.py", line 1998, in post_fp_vasp all_sys.append(_sys) File "/public/home/hfcas_user39/.local/lib/python3.9/site-packages/dpdata/system.py", line 1390, in append if not System.append(self, system): File "/public/home/hfcas_user39/.local/lib/python3.9/site-packages/dpdata/system.py", line 342, in append raise RuntimeError('systems with inconsistent formula could not be append: %s v.s. %s' % (self.uniq_formula, system.uniq_formula)) RuntimeError: systems with inconsistent formula could not be append: C28W36Zr28 v.s. C29W36Zr28
C28W36Zr28 is the formula of a defect-free configuration. However C29W36Zr28 represents a self-interstitial-carbon-atom cell, which is the definite result of an NVT simulation. So physically speaking, is it okay to merge systems with inconsistent amount of atoms ?
PS : Actually I didn't use perturbed configurations even in Initial sets. Instead I also added some defect configurations into my initial set, and the deepmd-kit worked well . Does the error above mean that in DpGen, one has to make the atom numbers of all configurations conserved ?
Thanks a lot !