Dimer method issue

[mikejwaters]

Hi, I'm trying to use ASE's dimer method to find saddle points as part of a larger workflow. The dimer method should work on a closely spaced pair of images of the same system. Due to their close proximity, the optimization procedure should behave like a the optimization of a single image, i.e. a series of small displacements. No parallel image is treatment needed. It appears that somewhere in the ASE code, it's setting chemical symbols a second time which rightly causes VaspInteractive to throw an error. Is there a way I can have VaspInteractive ignore this chemical symbol update? The dimer method doesn't change the order atoms in the system, anyway.

I am using the master branch of ASE and v0.1.0rc0 of VaspInteractive

Here's the traceback:

Traceback (most recent call last):
  File "/home/mjwaters/NbTiZr_tests/OTiZrNb_MLIPs/barrier_learning/DFT_test_set/structures/random_occupancy/0/../../../delete_relax_ends_and_dimer_for_dft_test_set.py", line 525, in <module>
    dim_rlx.run(fmax = fmax, steps=max_dimer_steps)
  File "/home/mjwaters/software/ase/ase/optimize/optimize.py", line 430, in run
    return Dynamics.run(self, steps=steps)
  File "/home/mjwaters/software/ase/ase/optimize/optimize.py", line 275, in run
    for converged in Dynamics.irun(self, steps=steps):
  File "/home/mjwaters/software/ase/ase/optimize/optimize.py", line 246, in irun
    self.step()
  File "/home/mjwaters/software/ase/ase/mep/dimer.py", line 1047, in step
    f0tp = self.dimeratoms.get_projected_forces(r0t)
  File "/home/mjwaters/software/ase/ase/mep/dimer.py", line 708, in get_projected_forces
    forces = self.get_forces(real=True, pos=pos).copy()
  File "/home/mjwaters/software/ase/ase/mep/dimer.py", line 671, in get_forces
    forces = self.atoms.get_forces()
  File "/home/mjwaters/software/ase/ase/atoms.py", line 818, in get_forces
    forces = self._calc.get_forces(self)
  File "/home/mjwaters/software/ase/ase/calculators/abc.py", line 30, in get_forces
    return self.get_property('forces', atoms)
  File "/home/mjwaters/software/ase/ase/calculators/calculator.py", line 538, in get_property
    self.calculate(atoms, [name], system_changes)
  File "/home/mjwaters/.local/lib/python3.10/site-packages/vasp_interactive/vasp_interactive.py", line 738, in calculate
    raise NotImplementedError(
NotImplementedError: VaspInteractive does not support change of chemical formula. Please create a new calculator instance or use standard Vasp calculator
Trying to close the VASP stream but encountered error: 
Traceback (most recent call last):
  File "/home/mjwaters/.local/lib/python3.10/site-packages/vasp_interactive/vasp_interactive.py", line 335, in _txt_outstream
    out = open(txt, "a")
NameError: name 'open' is not defined

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/home/mjwaters/.local/lib/python3.10/site-packages/vasp_interactive/vasp_interactive.py", line 1015, in _force_kill_process
    self.close()
  File "/home/mjwaters/.local/lib/python3.10/site-packages/vasp_interactive/vasp_interactive.py", line 588, in close
    with self._txt_outstream() as out:
UnboundLocalError: local variable 'out' referenced before assignment

Will now force closing the VASP process. The OUTCAR and vasprun.xml outputs may be incomplete
/home/mjwaters/.local/lib/python3.10/site-packages/vasp_interactive/vasp_interactive.py:621: UserWarning: VASP process no longer exists. Will reset the calculator.

Thanks,
-Mike

[yinkaaiwu]

Hi, did you solve this problem?

[mikejwaters]

I did not solve this problem. I can supply a lightweight example if it helps.

[yinkaaiwu]

I did not solve this problem. I can supply a lightweight example if it helps.

I didn't solve it either, I even tried the vaspinteractive calculator in ase and made some modifications to make it conform to ase.dimer's calculation flow, frustrating that it also didn't work. I feel that it may be necessary to rebuild both the calculator and the optimizer in order to migrate the dimer calculation process to vaspinteractive.

[mikejwaters]

This error doesn't make much sense.
NotImplementedError: VaspInteractive does not support change of chemical formula. Please create a new calculator instance or use standard Vasp calculator
There's no reason for the dimer method to change chemical formulas. This could be an ASE bug.
However, I cannot find any .get_* .set_* method that is touching the chemical symbols. I think that vaspinteractive might be behaving correctly.