The nep.in input file

From GPUMD
Revision as of 17:10, 6 May 2021 by Brucefan (talk | contribs) (An example)
Jump to navigation Jump to search

Brief descriptions

  • This file specifies some hyper-parameters of the NEP potential.

File format

  • This file has the following fixed form:
cutoff              [cutoff]
n_max               [n_max]
l_max               [l_max] 
num_neurons1        [num_neurons1] 
num_neurons2        [num_neurons2]
batch_size          [batch_size]
maximum_generation  [maximum_generation]
  • Do not modify the first column above.
  • For each row, items within [ ] are to be filled by the user.

An example

cutoff               4.0
n_max                6
l_max                4
num_neurons1         20
num_neurons2         0
batch_size           512
maximum_generation   100000
  • Explanations:
    • The cutoff distance of the potential is 4.0 A.
    • The expansion order for the radial part is [math]n_{\rm max}=6[/math].
    • The expansion order for the angular part is [math]l_{\rm max}=4[/math].
    • The number of neurons in the first hidden layer is 20.
    • The number of neurons in the second hidden layer is 0. Therefore, we are using a single hidden layer here.
    • The number of configurations used in one batch at one step is 512.
    • The training will be carried out for [math] 10^5[/math] generations.