Example 8, Thermal conductivity calculations with the SSCHA crashed

[yusufshaidu]

I am running example 6 tutorial the shows how to compute thermal conductivity with MLP but keep crashing at the first step.

Here is the error message:
"lattice = np.array(np.transpose(cell[0]), dtype="double", order="C")
TypeError: float() argument must be a string or a real number, not 'Atom'
"
The error seems to originate from "compute_phonons_finite_displacements_sym" called through compute_phonons_finite_displacements from the function "get_starting_dynamical_matrices". Does anyone knows the source of such errors?

Thanks!
Yusuf

[mesonepigreco]

Hi Yusuf,
Can you past here the full error stack?

[yusufshaidu]

Here is the full output.

/pscratch/sd/y/yshaidu/Code/sscph_env/lib/python3.10/site-packages/sscha/aiida_ensemble.py:18: UserWarning: aiida or aiida-quantumespresso are not installed
warnings.warn('aiida or aiida-quantumespresso are not installed')
Step[ FC] Time Energy fmax
Force-consistent energies used in optimization.
BFGSLineSearch: 0[ 0] 17:09:20 32.366892 174.7496
BFGSLineSearch: 1[ 1] 17:09:21 -7.695738* 10.9365
BFGSLineSearch: 2[ 3] 17:09:21 -9.073873* 3.2856
BFGSLineSearch: 3[ 4] 17:09:21 -9.246568* 0.8048
BFGSLineSearch: 4[ 5] 17:09:21 -9.259009* 0.0943
BFGSLineSearch: 5[ 6] 17:09:21 -9.259190* 0.0033
BFGSLineSearch: 6[ 7] 17:09:21 -9.259190* 0.0000
Traceback (most recent call last):
File "/pscratch/sd/y/yshaidu/lattice_thermal_conductivity/sscha/example_run1.py", line 49, in
dyn, bulk = get_starting_dynamical_matrices(
File "/pscratch/sd/y/yshaidu/lattice_thermal_conductivity/sscha/example_run1.py", line 28, in get_starting_dynamical_matrices
dyn = CC.Phonons.compute_phonons_finite_displacements(structure,
File "/pscratch/sd/y/yshaidu/Code/sscph_env/lib/python3.10/site-packages/cellconstructor/Phonons.py", line 4498, in compute_phonons_finite_displacements
return compute_phonons_finite_displacements_sym(structure, ase_calculator, epsilon,
File "/pscratch/sd/y/yshaidu/Code/sscph_env/lib/python3.10/site-packages/cellconstructor/Phonons.py", line 4675, in compute_phonons_finite_displacements_sym
symm = spglib.get_symmetry(super_structure.get_ase_atoms())
File "/pscratch/sd/y/yshaidu/Code/sscph_env/lib/python3.10/site-packages/spglib/spglib.py", line 248, in get_symmetry
lattice, _, _, magmoms = _expand_cell(cell)
File "/pscratch/sd/y/yshaidu/Code/sscph_env/lib/python3.10/site-packages/spglib/spglib.py", line 1985, in _expand_cell
lattice = np.array(np.transpose(cell[0]), dtype="double", order="C")
TypeError: float() argument must be a string or a real number, not 'Atom'

[diegomartinez2]

So this is the thermal conductivity example that takes the data from example 6. The error call is in spglib.py and comes from the structure data calculated in example 6 and read by ASE. This error may come from the original structure data if instead of the type of atom it says simply "Atom".
Can you show us your '../06_the_SSCHA_with_machine_learning_potentials/POSCAR' file?

[mesonepigreco]

It seems the problem is much bigger.
Togo just removed support for ase atoms in spglib, thus breaking all the compatibility of our code with spglib.

https://spglib.readthedocs.io/en/latest/releases.html

For now, the solution is to downgrade spglib prior 2.3

pip remove spglib
pip install spglib<=2.2 --force-reinstall

[yusufshaidu]

Thanks you very much for your help. This problem has been solved! But I encountered another issues.

/pscratch/sd/y/yshaidu/Code/sscph_env/lib/python3.10/site-packages/sscha/aiida_ensemble.py:18: UserWarning: aiida or aiida-quantumespresso are not installed
warnings.warn('aiida or aiida-quantumespresso are not installed')
Step[ FC] Time Energy fmax
Force-consistent energies used in optimization.
BFGSLineSearch: 0[ 0] 12:05:24 32.366892 174.7496
BFGSLineSearch: 1[ 1] 12:05:24 -7.695738* 10.9365
BFGSLineSearch: 2[ 3] 12:05:24 -9.073873* 3.2856
BFGSLineSearch: 3[ 4] 12:05:24 -9.246568* 0.8048
BFGSLineSearch: 4[ 5] 12:05:24 -9.259009* 0.0943
BFGSLineSearch: 5[ 6] 12:05:24 -9.259190* 0.0033
BFGSLineSearch: 6[ 7] 12:05:24 -9.259190* 0.0000
Number of symmetry inequivalent displacements: 1
Traceback (most recent call last):
File "/pscratch/sd/y/yshaidu/lattice_thermal_conductivity/sscha/example_run1.py", line 66, in
relax.vc_relax(static_bulk_modulus = bulk,
File "/pscratch/sd/y/yshaidu/Code/sscph_env/lib/python3.10/site-packages/sscha/Relax.py", line 604, in vc_relax
self.minim.ensemble.compute_ensemble(
File "/pscratch/sd/y/yshaidu/Code/sscph_env/lib/python3.10/site-packages/sscha/Ensemble.py", line 3848, in compute_ensemble
computing_ensemble = self.get_noncomputed()
File "/pscratch/sd/y/yshaidu/Code/sscph_env/lib/python3.10/site-packages/sscha/Ensemble.py", line 3989, in get_noncomputed
return self.split(non_mask)
File "/pscratch/sd/y/yshaidu/Code/sscph_env/lib/python3.10/site-packages/sscha/Ensemble.py", line 3942, in split
ens.update_weights(self.current_dyn, self.current_T)
File "/pscratch/sd/y/yshaidu/Code/sscph_env/lib/python3.10/site-packages/sscha/Ensemble.py", line 1950, in update_weights
old_a = SCHAModules.thermodynamic.w_to_a(w, self.T0)
AttributeError: module 'SCHAModules' has no attribute 'thermodynamic'

I am not sure what I doing wrong because I see that there is a file in SCHAModules that contains thermodynamic.

[mesonepigreco]

Glad it solved the issue.
Regarding your second error, there is the solution in #184
We are going to upgrade the code in the next days to fix these two problems

[yusufshaidu]

Thank you very much. I look forward to the upgrade!

In the meantime, I am now able to run the example but needed to change a little bit the new Thermodynamic python function. Basically, for instance, w_to_a needs not to take n as input as the shape of a could be infer from the shape of w (I think?) and also because n is not an argument of w_to_a in Ensemble.py:

I replaced
def w_to_a(w, T, n):
a = np.zeros(n)
with
def w_to_a(w, T):
a = np.zeros(w.shape)
and the other function that request n.

Please let me know if there is any problem with these changes.

Many thanks.

[mesonepigreco]

No correct, You are right, there is a mistake in that answer. I will update the accepted answer Il Sab 3 Feb 2024, 01:16 yusufshaidu ***@***.***> ha scritto:

…

Thank you very much. I look forward to the upgrade! In the meantime, I am now able to run the example but needed to change a little bit the new Thermodynamic python function. Basically, for instance, w_to_a needs not to take n as input as the shape of a could be infer from the shape of w (I think?) and also because n is not an argument of w_to_a in Ensemble.py: I replaced def w_to_a(w, T, n): a = np.zeros(n) with def w_to_a(w, T): a = np.zeros(w.shape) and the other function that request n. Please let me know if there is any problem with these changes. Many thanks. — Reply to this email directly, view it on GitHub <#186 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AGZMNVFX67GOQIGAZXXPR63YRV6XBAVCNFSM6AAAAABCV334W2VHI2DSMVQWIX3LMV43SRDJONRXK43TNFXW4Q3PNVWWK3TUHM4DGNJRHA3DS> . You are receiving this because you commented.Message ID: ***@***.***>