SpongeExt 1.3

Sponge Extension For Enhanced Sampling

SpongeExt

为MindSponge和CudaSPONGE设计的接口函数。CudaSPONGE部分通过SpongeExt，可以外接如PySAGES等基于Python的增强采样软件。MindSponge部分通过SpongeExt，可以扩展PyTorch后端，在GPU环境运行。

安装教程

从源码安装：

$ git clone https://gitee.com/dechin/sponge-ext.git && cd sponge-ext && python3 setup.py install --user

从pip安装：

$ python3 -m pip install SpongeExt --upgrade -i https://pypi.org/simple

CudaSPONGE使用案例

准备一系列的CudaSPONGE输入文件，例如examples中的nvt.in配置输入文件：

case1 MD simulation

mode = NVT
default_in_file_prefix = ../protein/alad

pbc=0 
cutoff=999

dt = 1e-3
step_limit = 2000
write_information_interval = 10

thermostat = middle_langevin
middle_langevin_gamma = 10

rst = nvt_restart

coordinate_in_file = ../protein/alad_coordinate.txt
plugin = /usr/local/python-3.7.5/lib/python3.7/site-packages/prips/_prips.so
py = ../pysages_metad.py

这里使用到了一个使用pysages定义的增强采样方法pysages_metad.py，内容为：

from SpongeExt import enhanced_sponge

import pysages
from pysages.colvars import DihedralAngle
from pysages.methods import Metadynamics

def phi_psi():
    from numpy import pi
    cvs = [DihedralAngle([4, 6, 8, 14]), DihedralAngle([6, 8, 14, 16])]
    height = 5.0  # kJ/mol
    sigma = [0.4, 0.4]  # radians
    stride = 3
    ngauss = 500
    grid = pysages.Grid(lower=(-pi, -pi), upper=(pi, pi), shape=(50, 50), periodic=True)
    method = Metadynamics(cvs, height, sigma, stride, ngauss, grid=grid)
    return method

pysages_method = phi_psi()
Calculate_Force, Mdout_Print = enhanced_sponge(pysages_method)

然后使用运行指令运行即可：

$ SPONGE -mdin nvt.in

输出内容大致如下图所示：

------------------------------------------------------------------------------------------------------------
           step =            1990,            time =           1.990,     temperature =         2388.79, 
      potential =           56.45,            CV_0 =       -1.500586,            CV_1 =       -2.550561, 
             LJ =           -0.79,         Coulomb =          -65.33,         nb14_LJ =            7.42, 
        nb14_EE =           41.19,            bond =            6.05,           angle =           40.54, 
       dihedral =           27.37, 
------------------------------------------------------------------------------------------------------------
           step =            2000,            time =           2.000,     temperature =         2430.97, 
      potential =           60.92,            CV_0 =       -0.989361,            CV_1 =       -3.084765, 
             LJ =           -0.22,         Coulomb =          -66.38,         nb14_LJ =            4.81, 
        nb14_EE =           42.10,            bond =            9.91,           angle =           38.81, 
       dihedral =           31.90, 
------------------------------------------------------------------------------------------------------------

配置深度学习后端

例如配置pytorch后端：

$ export DEEP_LEARNING_BACKEND=pytorch