ai2_kit.tool.model_devi module

ai2_kit.tool.model_devi module#

class ai2_kit.tool.model_devi.ModelDevi[source]#

Bases: object

A tool to analyze the deviation of model from model deviation file of deepmd-kit

done()[source]#
dump_stats(out_file: str = '', fmt='tsv')[source]#

Dump the statistics of grading

Parameters:

  • out_file – the file path to write the statistics

  • fmt – the format of table, default is tsv

grade(lo: float, hi: float, col: str = 'max_devi_f', outlier: Optional[float] = None)[source]#

Grade atoms based on the deviation of model into 3 levels: good, decent, poor

the grade is based on the column of max_devi_f by default, if the value is below lo, the level is good, if the value is above hi, the level is poor, otherwise, the level is decent. outlier is an independent indicator, default is 2 * hi if not provided.

Parameters:

  • lo – the lower bound of decent level, below this value is good

  • hi – the upper bound of decent level, above this value is poor

  • col – the column of model deviation to grade, default is max_devi_f

  • outlier – the threshold of outlier, default is 2 * hi

read(*dir_or_glob: str, traj_file: str, md_file='model_devi.out', ignore_error=False, **kwargs)[source]#

read model deviation from file, support multiple files and glob pattern

Parameters:

  • dir_or_glob – path or glob pattern to locate data path

  • traj_file – trajectory file name to read, relative to data path, e.g dump.lammpstrj

  • md_file – model deviation file name to read, default is model_devi.out

  • ignore_error – ignore error when reading files

  • kwargs – other arguments for ase.io.read

slice(expr: str)[source]#

Slice the atoms based on the expression

Parameters:

expr – the expression to slice the atoms, e.g. ‘0:10’

write(file_path: str, inplace=False, level='decent', ignore_error=False, **kwargs)[source]#

Write atoms to file based on the level of grading

Parameters:

  • file_path – the file path to write

  • inplace – if True, write to the original data directory

  • level – the level of grading to write, default is decent

  • ignore_error – ignore error when writing files

  • kwargs – other arguments for ase.io.write