Error in assigning pseudopotentials to the builder object

ansonthms · September 23, 2025, 11:25am

Dear all,

I am a new user of AiiDA, and was trying the example provided at 1. Quantum ESPRESSO — AiiDA Tutorials documentation

My issue is as follows:
After initializing the builder, attaching the structure element, and getting pseudos i.e.,

builder.structure = structure
pseudos = pseudo_family.get_pseudos(structure=structure)
pseudos

When I try to assign the pseudopotentials using

builder.pseudos = pseudos

The pseudopotentials are not being assigned.

Here is what I did:

In [12]: code = load_code(108)
...: builder=code.get_builder()
...: structure=load_node(109)
...: builder.structure=structure
...: pseudo_family = load_group('SSSP/1.3/PBE/efficiency')
In [13]: pseudos = pseudo_family.get_pseudos(structure=structure)
In [14]: builder.pseudos = pseudos
In [15]: pseudos
Out[15]: {'Si': <UpfData: uuid: d1e944a9-c320-41a2-b486-78f0ff7e759c (pk: 17)>}

In [16]: builder
Out[16]:
Process class: PwCalculation
Inputs:
code: qe-7.2@hpc2
metadata:
options:
stash: {}
monitors: {}
pseudos:
Si: ''
structure: Si

Suspecting this to be an issue with either the aiida-quantumespresso or the aiida-pseudo package, I have tried to reinstall both the packages from the latest version available at GitHub, but the issue still persists.

Further, the tests (using aiida-quantumespresso) related to PW also fail
================================= short test summary info ==================================
FAILED tests/calculations/test_pw.py::test_pw_validate_inputs_restart_base[scf]- TypeError: exceptions must be derived from Warning, not <class 'NoneType'>
FAILED tests/calculations/test_pw.py::test_pw_validate_inputs_restart_base[relax]- TypeError: exceptions must be derived from Warning, not <class 'NoneType'>
FAILED tests/calculations/test_pw.py::test_pw_validate_inputs_restart_base[vc-relax]- TypeError: exceptions must be derived from Warning, not <class 'NoneType'>
FAILED tests/workflows/protocols/pw/test_base.py::test_overrides_key_check[overrides0-None]- TypeError: exceptions must be derived from Warning, not <class 'NoneType'>
FAILED tests/workflows/protocols/pw/test_base.py::test_overrides_key_check[overrides1-None]- TypeError: exceptions must be derived from Warning, not <class 'NoneType'>
FAILED tests/workflows/protocols/pw/test_base.py::test_overrides_key_check[overrides2-None]- TypeError: exceptions must be derived from Warning, not <class 'NoneType'>
FAILED tests/workflows/protocols/pw/test_base.py::test_overrides_key_check[overrides3-None]- TypeError: exceptions must be derived from Warning, not <class 'NoneType'>
FAILED tests/workflows/protocols/pw/test_base.py::test_overrides_key_check[overrides4-None]- TypeError: exceptions must be derived from Warning, not <class 'NoneType'>
FAILED tests/workflows/protocols/pw/test_base.py::test_overrides_key_check[overrides5-None]- TypeError: exceptions must be derived from Warning, not <class 'NoneType'>
FAILED tests/workflows/protocols/pw/test_relax.py::test_overrides_key_check[overrides0-None]- TypeError: exceptions must be derived from Warning, not <class 'NoneType'>
FAILED tests/workflows/protocols/pw/test_relax.py::test_overrides_key_check[overrides1-None]- TypeError: exceptions must be derived from Warning, not <class 'NoneType'>
FAILED tests/workflows/protocols/pw/test_relax.py::test_overrides_key_check[overrides2-None]- TypeError: exceptions must be derived from Warning, not <class 'NoneType'>
============ 12 failed, 525 passed, 2 skipped, 43 warnings in 542.66s (0:09:02) ============

Could someone suggest how to fix this?

mbercx · September 23, 2025, 11:35am

Welcome @ansonthms! I fear the issue is in the representation of the builder in IPython/Jupyter notebooks. Try just running

builder.pseudos

And you’ll probably get the same output as before.

More technically: the __repr_pretty__ method of the ProcessBuilder class uses a custom JSONEncoder to print the builder contents, that relies on the get_description method of the nodes. But we haven’t implemented that yet for the pseudo potential nodes in aiida-pseudo. I’ll open an issue to track this.

mbercx · September 23, 2025, 11:40am

Be careful with aiida-quantumespresso here! We’re currently working on a new major release on main, so there might be breaking changes.

github.com/aiidateam/aiida-pseudo

👌 Add `get_description` method to pseudo potential classes

opened 11:39AM - 23 Sep 25 UTC

mbercx

As a user notes on Discourse: https://aiida.discourse.group/t/error-in-assigning…-pseudopotentials-to-the-builder-object/681/2 the representation of a `ProcessBuilder` class in IPython/Jupyter notebooks seems to contain no pseudos: ```python Process class: PwCalculation Inputs: code: qe-7.2@hpc2 metadata: options: stash: {} monitors: {} pseudos: Si: '' structure: Si ``` But that is because the [__repr_pretty__ method](https://github.com/aiidateam/aiida-core/blob/7255f01c7e4e12d65d2a7ffa356552b737c58cb3/src/aiida/engine/processes/builder.py#L242-L249) of the ProcessBuilder class uses a [custom JSONEncoder](https://github.com/aiidateam/aiida-core/blob/7255f01c7e4e12d65d2a7ffa356552b737c58cb3/src/aiida/engine/processes/builder.py#L28-L39) to print the builder contents, that relies on the get_description method of the nodes. But we haven’t implemented that yet for the pseudo potential nodes here. Time to change that, it's clearly confusing, and not a lot of work.

ansonthms · September 23, 2025, 11:41am

Thank you @mbercx for your response.

Yes, builder.pseudos does respond correctly, however, the when printing out of just builder, it is not assigned.

In [17]: builder.pseudos
Out[17]: {‘Si’: <UpfData: uuid: d1e944a9-c320-41a2-b486-78f0ff7e759c (pk: 17)>}

In [18]: builder
Out[18]:
Process class: PwCalculation
Inputs:
code: qe-7.2@hpc2
metadata:
options:
stash: {}
monitors: {}
pseudos:
Si: ‘’
structure: Si

mbercx · September 23, 2025, 11:44am

As I explained, that’s just an issue with how the builder is printed. I assure you the pseudos are there.

I’ll ping you when I have a fix, so you can try the new aiida-pseudo branch and give feedback. It’s 9:45 pm where I am though (AEST), so that’ll likely be for tomorrow.

ansonthms · September 23, 2025, 11:48am

Much appreciated. Thank you so much

mbercx · September 23, 2025, 2:06pm

github.com/aiidateam/aiida-pseudo

👌 `PseudoPotentialData`: Improve string representation

main ← imp/pseudo-description

opened 02:03PM - 23 Sep 25 UTC

mbercx

+7 -0

The `PseudoPotentialData` is currently missing a `get_description` method, which… means this returns the default empty string derived from the `Node` class. The default `__repr__` method also does not show one of the most important attributes of the `PseudoPotentialData` class: the element it represents. Here we add the `__repr__` and `get_description` methods to the `PseudoPotentialData` class. Although these could (and perhaps should) serve a different purpose, we don't want users to be confused when they see a difference between: * the pseudos dictionary returned by `builder.pseudos`, which uses `__repr__`. * The `__repr_pretty__` method of the `ProcessBuilder`, which uses `get_description` for `Node` objects. Hence, we make them return the same output by having the `get_description` method call `repr`. For the contents, we choose to add: * The class name, so they can see the type of pseudopotential. * The element of the pseudopotential. * The PK of the pseudpotential node. We skip the UUID to reduce the noise. In most local usage the PK is much more common, and the user can still easily get the UUID using `node.uuid`.

Example usage from aiida-quantumespresso:

from aiida import orm, load_profile 

load_profile()

from ase.build import bulk
from aiida_quantumespresso.calculations.pw import PwCalculation

structure = orm.StructureData(ase=bulk('Si', 'fcc', 5.43))
pseudo_family = orm.load_group('SSSP/1.3/PBEsol/efficiency')

builder = PwCalculation.get_builder()
builder.pseudos = pseudo_family.get_pseudos(structure=structure)
builder

returns

Process class: PwCalculation
Inputs:
metadata:
  options:
    stash: {}
    unstash: {}
monitors: {}
pseudos:
  Si: '<UpfData[Si] (pk: 72)>'

Note that I also adapt the __repr__, so

builder.pseudos

returns

{'Si': <UpfData[Si] (pk: 72)>}

Note: released in v1.8.0

system · October 3, 2025, 4:24am

This topic was automatically closed 5 days after the last reply. New replies are no longer allowed.

Topic		Replies	Views
Pseudopotentials in aiida workflow example with Espresso General Usage	3	127	September 17, 2025
Aiida_vibroscopy related question: ValueError: required pseudo family `SSSP/1.3/PBEsol/efficiency` is not installed. Please use `aiida-pseudo install` toinstall it New to AiiDA	12	139	February 14, 2024
New to AiiDA. Question about loading specific pseudo families New to AiiDA	2	80	July 23, 2024
Cannot load ONCV pseudopotential contains pseudo potentials with duplicate elements New to AiiDA	2	223	February 28, 2024
aiida.engine.processes.calcjobs.tasks.PreSubmitException: exception occurred in presubmit call New to AiiDA	2	113	February 2, 2024

Error in assigning pseudopotentials to the builder object

Related topics