Implemented PBC and ECP. #160
Conversation
|
I have switched this request to Draft because it was ok with |
|
Dear @sunqm, I could not reproduce the check failure (Only the |
…k_gamma_ae_6_31g` and `test_cell_k_grid_ae_6_31g`. `int1e_nuc` checks randomly fail in `test_cell_k_gamma_ae_6_31g` and `test_cell_k_grid_ae_6_31g`. We confirm that the internal values are consistent between `cell0` and `cell1`. So, it does not seem due to a TREXIO read/write error.
| kpt = mol.get_scaled_kpts(kpt) | ||
| if np.all(np.array(kpt) == 0.0): | ||
| # gamma point PBC. Real wavefunction | ||
| kpts = np.array(mf.kpts) |
There was a problem hiding this comment.
k-point calculations can be created with a single k-point, such as mf = cell.KRKS(xc='pbe'). This case will go through the nk == 1 branch, which assumes the orbital coefficients a 2D array. This might be problematic. mf.mo_coeff here is an array with shape (1, N, N)
There was a problem hiding this comment.
@sunqm, Thank you very much for the review. Yes, this is a corner case that I have overlooked. I have fixed it and introduced corresponding new tests test_mf_k_single_grid_rhf_ae_6_31g and test_mf_k_single_grid_uhf_ae_6_31g. They failed with the previous version of trexio.py, while they pass right now.
…int) that I have overlooked. Introduced corresponding new tests.
2 participants
Implemented the following functions in
pyscf/tools/trexio.py. See also #153.[Writing]
_scf_to_trexio: Complex wavefunctions and twisted-average are done.[Reading]
mol_from_trexio: The ECP implementationif trexio.has_ecp(tf)is done.scf_from_trexio: UHF and ROHF are also supported.[Test suites]
trexiotool.