An error occurred when using graph_data_gen

[Zhengsx-p]

Dear Yang Zhong,
I want to uss HamEPC to calculate the mobility of CsPbBr3, and a HamGNN model has been trained. I used perturb_supercell.py to generate the supercells and openmx_postprocess to obtain overlap.scfout. However, when using graph_data_gen to obtain graph_data.npz of those supercells and setting nao_max to 26, I received the following error message:

Traceback (most recent call last):
  File "/lustre/home/2401110483/software/miniconda3/envs/HamGNN/bin/graph_data_gen", line 33, in <module>
    sys.exit(load_entry_point('HamGNN==2.0.0', 'console_scripts', 'graph_data_gen')())
  File "/lustre/home/2401110483/software/miniconda3/envs/HamGNN/lib/python3.9/site-packages/HamGNN-2.0.0-py3.9.egg/utils_openmx/graph_data_gen.py", line 307, in main
    H[i][mask] = np.array(sub_maxtrix_H)
ValueError: NumPy boolean array indexing assignment cannot assign 361 input values to the 676 output values where the mask is true

but if I set nao_max as 19, there is no definition of Br in def_basis_19 and I received the following error message:

Traceback (most recent call last):
  File "/lustre/home/2401110483/software/miniconda3/envs/HamGNN/bin/graph_data_gen", line 33, in <module>
    sys.exit(load_entry_point('HamGNN==2.0.0', 'console_scripts', 'graph_data_gen')())
  File "/lustre/home/2401110483/software/miniconda3/envs/HamGNN/lib/python3.9/site-packages/HamGNN-2.0.0-py3.9.egg/utils_openmx/graph_data_gen.py", line 305, in main
    mask[basis_def[src][:,None], basis_def[src][None,:]] = 1
KeyError: 35

Besides, when I trained the HamGNN model and set nao_max to 26, there was no error.
Could you please tell me what is causing this error and how I should resolve it?

Thanks
Shixuan Zheng

[QuantumLab-ZY]

Hi Shixuan Zheng,

When generating the graph_data.npz, it's essential to ensure that the nao_max parameter is consistent across both the graph_data_gen.yaml and config.yaml files.

Regarding the error you encountered with CsPbBr3 when setting nao_max to 26, it is likely due to a mismatch in the atomic orbital basis sets used in your OpenMX calculations. Make sure you are using the following basis sets:

• Cs: Cs12.0-s3p2d2
• Pb: Pb8.0-s3p2d2f1
• Br: Br7.0-s3p2d2

Any deviation from these specified basis sets could lead to the errors you're seeing. Please check your configuration and adjust as necessary.

Best regards,
Yang Zhong

[Zhengsx-p]

Dear Yang Zhong,
Thank you for your helpful response. After looking into it more carefully, I noticed that in HamEPC/grad_mat_prep/utils.py, 'PAO_dict' had 'Pb':'Pb8.0-s3p2d2' set, which differs from my setting in HamGNN training. I resolved the issue by changing it to "Pb": "Pb8.0-s3p2d2f1". I'm grateful for your assistance!

Best wishes,
Shixuan Zheng