dpdata: RuntimeError: system has virials, but this does not

[MaterialDesign-USTS]

I used dpdata to prepare training_data and validation_data from a series OUTCAR of AIMD. However, the dpdata stopped due to the error:
RuntimeError Traceback (most recent call last)
Cell In[1], line 50
48 data.append(data_6ring1)
49 data.append(data_6ring2)
---> 50 data.append(data_10ring1_shxy01)
51 data.append(data_10ring1_shxz01)
52 data.append(data_10ring1_shyx01)

File /opt/deepmd-kit-3.1.0/lib/python3.12/site-packages/dpdata/system.py:1509, in MultiSystems.append(self, *systems)
1507 for system in systems:
1508 if isinstance(system, System):
-> 1509 self.__append(system)
1510 elif isinstance(system, MultiSystems):
1511 for sys in system:

File /opt/deepmd-kit-3.1.0/lib/python3.12/site-packages/dpdata/system.py:1524, in MultiSystems.__append(self, system)
1522 formula = system.formula
1523 if formula in self.systems:
-> 1524 self.systems[formula].append(system)
1525 else:
1526 self.systems[formula] = system.copy()

File /opt/deepmd-kit-3.1.0/lib/python3.12/site-packages/dpdata/system.py:507, in System.append(self, system)
505 if tt.shape is not None and Axis.NFRAMES in tt.shape:
506 if tt.name not in self.data and tt.name in system.data:
--> 507 raise RuntimeError(f"system has {tt.name}, but this does not")
508 elif tt.name in self.data and tt.name not in system.data:
509 raise RuntimeError(f"this has {tt.name}, but system does not")

RuntimeError: system has virials, but this does not
How to fix it?

[dosubot[bot]]

The error happens because some of your systems have virials (stress tensor data) and others do not, and dpdata requires all systems in a MultiSystems object to have the same set of data fields. This usually occurs if some OUTCAR files are missing virial data, often due to VASP INCAR settings like NWRITE not being set to 2, or due to parsing issues if the OUTCAR format is inconsistent (e.g., extra lines like vdW contributions) [source] [source] [source].

To fix:

1. Check which systems have virials:
   Use the has_virial() method in dpdata to check each system:

   for sys in [data_6ring1, data_6ring2, data_10ring1_shxy01, ...]:
       print(sys.has_virial())

   All should return the same value (all True or all False).

2. Ensure consistent virial data:

   • If you need virials for training, regenerate any missing virials by rerunning AIMD with NWRITE=2 in your VASP INCAR file [source].
   • If you do not need virials, you can remove the virials field from all systems before appending:

     for sys in [data_6ring1, data_6ring2, data_10ring1_shxy01, ...]:
         if sys.has_virial():
             del sys.data["virials"]

   • If the issue is due to parsing (e.g., OUTCAR format differences), check that dpdata is correctly extracting virials. Sometimes, extra lines (like vdW) shift the expected location of the stress/virial data. You may need to preprocess your OUTCAR files or adjust dpdata's parsing logic [source].

3. Upgrade dpdata:
   Make sure you are using the latest version of dpdata, as recent releases have fixed several related bugs.

Best practice: always ensure all systems you append have the same set of data fields (virials, forces, energies, etc.) to avoid this and similar errors.

To reply, just mention @dosu.

---

Docs are dead. Just use Dosu.

[njzjz-bot]

The main diagnosis in Dosu's answer is correct: systems being appended must consistently either contain virials or omit them. However, NWRITE=2 is not a general requirement for VASP virial/stress output and should not be assumed to be the fix.

Check each parsed system directly:

for name, system in systems:
    print(name, "virials" in system.data, system.get_nframes())

A missing virial commonly indicates an incomplete/unconverged/truncated OUTCAR, a differently generated calculation, or a parser issue. Regenerate or reparse those calculations so every frame has consistent energy, force, and virial labels if you intend to train stresses.

If virials are intentionally not needed, remove the field consistently from all systems before appending:

for system in systems:
    system.data.pop("virials", None)

Do not fill missing virials with zeros: that creates incorrect stress labels. It is also useful to load and append one OUTCAR at a time so the first failing filename is explicit, then inspect whether its ionic/electronic calculation completed normally.

Coding agent: Codex
Codex version: codex-cli 0.149.0
Model: gpt-5.6-sol
Reasoning effort: xhigh