A lightweight Python 3.7+ package for reading and writing VASP ML_AB files.

Clone the repository and install using

pip install .

Pymlff depends on numpy and click.

Pymlff includes a basic command line utility for merging ML_AB.

mlff merge ML_AB1 ML_AB2 ML_AB_NEW

More than 2 ML_AB files can be combined at once, e.g.

mlff merge ML_AB1 ML_AB2 ML_AB3 ML_AB4 ML_AB_NEW

Pymlff also includes a command line utility for converting ML_AB files to extended xyz (extxyz) files. The last argument is the units to use for converting the VASP kbar units. Valid units are eV/A^3 and kbar.

# Convert ML_AB to extxyz
mlff write-extxyz ML_AB ML_AB.xyz

# Convert ML_AB to extxyz and convert the units of stress from kbar to eV/A^3 (applies a negative sign)
mlff write-extxyz ML_AB ML_AB.xyz --stress-unit=eV/A^3

More functionality is available through the Python API.

from pymlff import MLAB

# load an ML_AB file
ab = MLAB.from_file("ML_AB")

# write an ML_AB file
ab.write_file("ML_AB_NEW")

Configurations can be filtered based on a filter function. The function should accept two arguments, the index of the configuration in the configurations list and a configuration itself. If the filter function returns True, the configuration will be kept, if False, the configuration will be removed. For example, to filter configurations with less than 256 atoms:

new_ab = ab.filter_configurations(lambda i, config: config.num_atoms >= 256)

See the Configuration object for the available attributes to use for filtering.

MLAB objects can be combined using the Python + operator. For example,

from pymlff import MLAB

# load two ML_AB files
ab1 = MLAB.from_file("ML_AB1")
ab2 = MLAB.from_file("ML_AB2")

new_ab = ab1 + ab2

MLAB objects can be written to the extxyz format using the write_extxyz method.

from pymlff import MLAB

# load an ML_AB file
ab = MLAB.from_file("ML_AB")

# write an extxyz file
ab.write_extxyz("ML_AB.xyz", "eV/A^3")