mmml pyscf-evaluate¶
Batch E/F/D/ESP evaluation.
Usage¶
mmml pyscf-evaluate --help
Options¶
usage: mmml pyscf-evaluate [-h] -i INPUT [-o OUTPUT] [--basis BASIS] [--xc XC]
[--spin SPIN] [--charge CHARGE] [--no-energy]
[--no-gradient] [--no-dipole] [--esp]
[--esp-cpu-fallback] [--polarizability] [--EF]
[--efield Ex,Ey,Ez] [--efield-sigma EFIELD_SIGMA]
[--add-random-noise SIGMA] [--seed SEED]
[--no-efield-include-nuclear-energy]
Evaluate geometries with pyscf-dft (energy, forces, dipoles, ESP).
Input & configuration:
-i, --input INPUT Input NPZ with R, Z, N (e.g. from normal-mode-sample)
Scientific model:
--basis BASIS Basis set (default: def2-SVP)
--xc XC XC functional (default: PBE0)
--spin SPIN 2*spin (0=singlet, 1=doublet, default: 0)
--charge CHARGE Total charge (default: 0)
--no-energy Skip energy (not recommended)
--EF Include uniform electric field in the Hamiltonian
(atomic units). Without --efield, draw a random
(Ex,Ey,Ez) per geometry (see --efield-sigma). Giving
--efield alone also enables the field (same vector for
all frames).
--efield Ex,Ey,Ez Fixed field in a.u., same for all geometries (enables
E-field even without --EF). If the first component is
negative, use equals form, e.g. --efield=-0.01,0,0
(argparse otherwise treats -0.01,... as a separate
flag).
--efield-sigma EFIELD_SIGMA
Std dev (a.u.) per component for random fields when --EF
is set without --efield (default: 0.01)
--no-efield-include-nuclear-energy
With --EF/--efield: use mf.kernel energy only (omit
nuclear-field term after SCF).
Execution:
--seed SEED RNG seed for --add-random-noise and random --EF draws
Output & artifacts:
-o, --output OUTPUT Output NPZ path (default: evaluated.npz)
Diagnostics & safety:
-h, --help show this help message and exit
Other options:
--no-gradient Skip forces/gradients
--no-dipole Skip dipole moments
--esp Compute ESP on density-selected grid (slower)
--esp-cpu-fallback Use CPU path for ESP (slower; default: GPU int1e_grids)
--polarizability Compute molecular polarizability tensor for each
geometry
--add-random-noise SIGMA
Gaussian noise std dev in Angstrom added to all R
components before evaluation
CLI to evaluate sampled geometries with pyscf-dft (energy, forces, dipoles,
ESP). Runs all geometries in one process (same GPU context) for speed. Input:
NPZ with R (n_samples, n_atoms, 3), Z, N (e.g. from normal-mode-sample) Output:
NPZ with R, Z, N, E, F, Dxyz, esp, esp_grid (if --esp), Ef (if --EF) Usage: mmml
pyscf-evaluate -i out/06_sampled.npz -o out/07_evaluated.npz mmml pyscf-evaluate
-i out/06_sampled.npz -o out/07_evaluated.npz --esp mmml pyscf-evaluate -i
traj.npz -o out.npz --EF mmml pyscf-evaluate -i traj.npz -o out.npz --EF
--efield 0,0,0.01 mmml pyscf-evaluate -i traj.npz -o out.npz --efield=-0.01,0,0
mmml pyscf-evaluate -i traj.npz -o out.npz --EF --no-efield-include-nuclear-
energy mmml pyscf-evaluate -i traj.npz -o out.npz --add-random-noise 0.1