convert

The dptools convert command is used to conveniently convert between different structure file types and concatenate multiple input files into a single output file. It is generally used as a convenient tool for setting up DFT calculation results to be used with the input command.

General usage,

$ dptools convert [-h] [-i INDICES] [-r REPEAT] input [input ...] output

positional arguments:
  input                 Input files (with extensions) to convert. Multiple inputs are concatenated into output
  output                Output file name to write conversion to (with extension)

optional arguments:
  -h, --help            show this help message and exit
  -i INDICES, --indices INDICES
                        Indices of input files to read. E.g., :10, -3:, :100:5 (default: :)
  -r REPEAT, --repeat REPEAT
                        Repeat each input before writing to output (e.g., 222) (default: None)

Quick reference examples

$ dptools convert thing.cif thing.traj
$ dptools convert md_000/vasprun.xml md_001/vasprun.xml full_md.traj
$ dptools convert md_???/vasprun.xml full_md.traj # equivalent to above
$ dptools convert 0*/vasp_spe.traj all_spe.traj
$ dptools convert -i ::10 vasprun.xml condensed_traj.traj

Simple file type conversion

The simplest usage of dptools convert is to convert from one structural file type to another. Such as converting the file atoms.cif to atoms.xyz,

$ dptools convert atoms.cif atoms.xyz

Concatenate MD trajectories

The convert command can also be used to concatenate separate MD trajectories into a single trajectory (especially useful if running variable time AIMD jobs that start and restart in a new directory). The final image in each input is compared with the first image of the previous input to ensure that only unique images are written.

If for example you have multiple vasprun.xml files from a single variable-time MD run located in md_000, md_001, …, md_xxx folders, you can concatenate them all into a single trajectory using this command (make sure to take advantage of Bash wildcards when using the command line!),

$ dptools convert md_???/vasprun.xml iteration0_system1.db

Note the unique name given to the output file. Naming outputs in this way makes it trivial to then setup a training set for deepmd-kit with input.

Concatenate single point energy calculations

Similar to the previous example, convert can be used to concatenate multiple single point energy calculations into a single file. This is useful if running single points from DPMD trajectories in separate directories to get more DFT data for training a new model iteration.

$ dptools convert 0*/vasp_spe.traj iteration1_system1.db

Again, notice the unique name assigned to the output!

Condense long MD run

If you have a long MD run that you want to condense by taking every 100th image, just include the appropriate index slice with the -i flag,

$ dptools convert -i ::100 vasprun.xml condensed_run.traj

Note

It is usually a good idea to condense long MD runs to avoid redundant data points in your training sets. Training on every image of a 1,000,000 step MD run is inefficient! You could get comparable results training on every 100th image without taking up tons of space on your hard drive.