Quickstart guide

A short example illustrating how to run the workflow directly from the command-line.

See also

• Python Examples

• Input

• Tutorial using the graphical user interface

Important

You must first install M3GNet before running this example!

# Example to train an M3GNet potential to reproduce the potential energy surface
# of the UFF force field for a small organic molecule

# In real applications, use a different reference engine!

# Before running this example you must install m3gnet:
# "$AMSBIN/amspackages" install m3gnet

# For details or off-line installation, see the package manager documentation.


"$AMSBIN/simple_active_learning" <<EOF
    ActiveLearning
      Steps
        Geometric
          NumSteps 5
          Start 10
        End
        Type Geometric
      End
    End

    MachineLearning
      Backend M3GNet
      CommitteeSize 1
      M3GNet
        Model UniversalPotential
      End
      MaxEpochs 200
    End

    MolecularDynamics
      InitialVelocities
        Temperature 300.0
        Type Random
      End
      NSteps 10000
      Thermostat
        Tau 200.0
        Temperature 300.0
        Type NHC
      End
      TimeStep 0.5
      Trajectory
        SamplingFreq 100
      End
    End

    Task MolecularDynamics

    Engine ForceField
      Type UFF
    EndEngine

    System
      Atoms
                  O       1.5185424677       1.2528427606      -0.3382346351
                  C       1.0167107700       0.2231953999       0.1069866215
                  C      -0.3341736669      -0.1931701775      -0.3199821682
                  O      -1.3067693409       0.4447398660       0.4319048572
                  H       1.6797677292      -0.4284512471       0.6676972986
                  H      -0.4715138766       0.0855120883      -1.3968084763
                  H      -0.4208687404      -1.2969328351      -0.3134514232
                  H      -1.6816953421      -0.0877358551       1.1618879254
      End
    End
EOF