ai2_kit.algorithm.reweighting module#
- class ai2_kit.algorithm.reweighting.ReweightingTool[source]#
Bases:
object- load_colvar(*path_or_glob: str)[source]#
load PLUMED COLVAR files and concatenate into a single DataFrame you have to ensure the COLVAR data is aligned with energies
- Parameters:
path_or_glob – path or glob pattern to locate data path
- load_energy(*path_or_glob: str, tag: str)[source]#
read baseline data as dpdata.LabeledSystem
- Parameters:
path_or_glob – path or glob pattern to locate data path
tag – a string tag to distinguish data, it is suggested to use baseline and target
- reweighting(cv: str, bias: str, temp: float, grid_size=0.01, save_fig_to: str | None = None, save_json_to: str | None = None, baseline_tag='baseline', target_tag='target')[source]#
run reweighting against loaded data
- Parameters:
cv – name of collective variable columns, for example d1
bias – name of bias column, for example opes.bias
temp – temperature
save_to – save th
- ai2_kit.algorithm.reweighting.compute_fes(cvs: ndarray, bias: ndarray, temp: float, grid=None, w=1, kB=0.0083144621, grid_size=100j)[source]#
Compute free energy surface from biased sampling data using Gaussian KDE Support 1D and 2D collective variables
- Parameters:
cv – collective variable
bias – bias potential
temp – temperature
grid – grid points, if None, will use grid_size to generate automatically
w – weights
kB – Boltzmann constant
grid_size – grid size in np.mgrid style
- Returns:
free energy
- ai2_kit.algorithm.reweighting.compute_kde_weight(baseline_energy: ndarray, target_energy: ndarray, temp: float, kB=0.0083144621, ev_to_kjmol=96.4853365)[source]#
Compute KDE weights for reweighting
- Parameters:
baseline_energy – baseline energy in eV, it’s the output of DeepMD evaluation
target_energy – target energy in eV, it’s the output of DeepMD evaluation