代码拉取完成,页面将自动刷新
import torch
from torch import Tensor
import numpy as np
import h5py
from torchsponge import Sponge
from torchsponge.potential import EnergyCell
from torchsponge.potential.pyscfcell import PySCFForceCell
from torchsponge import Protein, WithForceCell, UpdaterMD, ForceField, WithEnergyCell, RunOneStepCell, ForceFieldBase
from torchsponge.callback import WriteH5MD
from torchsponge.sampling import Metadynamics
from torchsponge.function import VelocityGenerator
from torchsponge.colvar import Torsion
from torchsponge.function import PI
mol = Protein(pdb='alad.pdb')
potential = ForceField(mol, 'AMBER.FF14SB')
phi = Torsion([4, 6, 8, 14])
psi = Torsion([6, 8, 14, 16])
min_opt = torch.optim.Adam(mol.parameters(), 1e-4)
mini = Sponge(mol, potential, min_opt, metrics={'phi': phi, 'psi': psi})
mini.run(100, 10)
class BlankEnergy(EnergyCell):
def __init__(self, name: str = 'energy', length_unit: str = 'nm', energy_unit: str = 'kj/mol', use_pbc: bool = None, **kwargs):
super().__init__(name=name, length_unit=length_unit, energy_unit=energy_unit, use_pbc=use_pbc, **kwargs)
def forward(self, coordinate: Tensor, neighbour_index: Tensor = None, neighbour_mask: Tensor = None,
neighbour_vector: Tensor = None, neighbour_distance: Tensor = None, pbc_box: Tensor = None):
b = coordinate.shape[0]
energy = torch.zeros(b, 1, dtype=torch.float32, device=coordinate.device)
return energy
metad = Metadynamics(
colvar=[phi, psi],
update_pace=10,
height=2.5,
sigma=0.05,
grid_min=-PI,
grid_max=PI,
grid_bin=360,
temperature=300,
bias_factor=100,
)
blank = BlankEnergy()
ff = ForceFieldBase(blank)
bs = WithEnergyCell(mol,ff,bias=metad)
pyscf_frc = PySCFForceCell(system=mol, pyscf_option='ks', mol_parameters={'basis': '6-31g','verbose':0}, ks_parameters={'xc':'b3lyp'}, use_pbc=False)
wf = WithForceCell(mol, pyscf_frc)
vgen = VelocityGenerator(300)
velocity = vgen(mol.shape, mol.atom_mass)
opt = UpdaterMD(
mol,
temperature = 300,
integrator='leap_frog',
thermostat='langevin',
time_step=2e-3,
velocity=velocity
)
onestep = RunOneStepCell(energy=bs, force=wf, optimizer=opt)
md = Sponge(network=onestep, write_h5md=WriteH5MD(mol, 'test.h5md', 100, write_metrics=True, write_bias=True), metrics={'phi': phi, 'psi': psi})
md.run(100,10,10) # total steps, print_interval, write_info_interval
with h5py.File('test.h5md') as f:
md_hills = np.array(f['parameters']['meta_potential']['hills0'])
md_grids = np.array(f['parameters']['meta_potential']['grids0'])
print(md_hills)
print(md_grids)
此处可能存在不合适展示的内容,页面不予展示。您可通过相关编辑功能自查并修改。
如您确认内容无涉及 不当用语 / 纯广告导流 / 暴力 / 低俗色情 / 侵权 / 盗版 / 虚假 / 无价值内容或违法国家有关法律法规的内容,可点击提交进行申诉,我们将尽快为您处理。