abacustest is mainly designed to handle pre-processing and post-processing for calculations performed by ABACUS. Moreover, it supports high-throughput submission of computing tasks to the Bohrium cloud platform and supercomputers for calculation.
- Install via pip:
pip install abacustest
- Or install from source code:
git clone https://github.com/pxlxingliang/abacus-test.git
cd abacus-test
pip install .
After installation, you can use the command-line tool abacustest.
Please use abacustest -h to get the usage instructions.
NOTE: in abacustest, an example is a folder that contains the ABACUS input files, such as INPUT, KPT, STRU, etc.
The prepare function of abacustest supports preparing input files for ABACUS. It requires a parameter configuration file in JSON format (e.g., param.json).
To initiate the input file preparation process, execute the command:
abacustest prepare -p param.json -s abacustest
-s specifies the directory where the input files will be stored. If not specified, it defaults to "abacustest".
The structure of param.json is as follows:
{
"prepare": {
"example_template": ["example1", "example2"],
"input_template": "INPUT",
...
}
}
Only the key "prepare" is recognized by abacustest prepare, and its value is a dictionary that defines the parameters required for preparing the input files. The parameters containing the following sections:
abacustest supports converting structure files into ABACUS input files. This is useful for preparing the STRU files needed for ABACUS calculations. \
The param.json may be like:
{
"prepare":{
"strus": ["1.vasp", "2.vasp"],
"stru_format": "poscar",
"input_template": "INPUT",
"kpt_template": "KPT",
"pp_dict": {"H": "H.upf", "O": "O.upf"},
"orb_dict": {"H": "H.orb", "O": "O.orb"},
"pp_path": "path/to/pp",
"orb_path": "path/to/orb"
}
}
The configuration file and its description are divided into the following sections:
-
strus/stru_format: Convert structure files fromstru_formatto ABACUS STRU format.strusshould be a string or list of strings representing paths readable by dpdata, andstru_formatshould be a string representing a format supported by dpdata. Common formats include "poscar" for POSCAR files and "deepmd/npy" for deepmd npy files. A special format is "cif" for CIF files, which will be parsed by ASE. The converted STRU files will be placed in new folders named 000000, 000001, 000002, etc. -
input_template: The template for the INPUT file. If specified, all examples will use this file as their INPUT file. -
kpt_template: The template for the KPT file. If specified, all examples will use this file as their KPT file. -
pp_dict: A dictionary of pseudopotentials. Keys are element names, and values are pseudopotential filenames. For example:{"H": "H.upf", "O": "O.upf"}. The related files will be written into STRU file, and create a soft link to the pseudopotential file in each example folder. -
orb_dict: A dictionary of orbitals. Keys are element names, and values are orbital filenames. For example:{"H": "H.orb", "O": "O.orb"}. The related files will be written into STRU file, and create a soft link to the orbital file in each example folder. -
pp_path: The path to pseudopotential library. An additional "element.json" JSON file should be present, defining element names and corresponding pseudopotential filenames (e.g.,{"H": "H.psp8", "O": "O.psp8"}). If not, abacustest will use the first two letters of pseudopotential filenames as element names. If an element is defined in bothpp_dictandpp_path, thepp_dictvalue takes precedence. -
orb_path: The path to orbital library. An additional "element.json" file should be present, defining element names and corresponding orbital filenames (e.g.,{"H": "H.orb", "O": "O.orb"}). If not, abacustest will use the first two letters of orbital filenames as element names. If an element is defined in bothorb_dictandorb_path, theorb_dictvalue takes precedence.
abacustest also supports preparing input files for ABACUS parameter testing. This is useful for generating multiple input files with different parameters for high-throughput calculations. \
The param.json may be like:
{
"prepare": {
"example_template": ["example1", "example2"],
"mix_input": {"ecutwfc": [50, 60, 70],
"kspacing": [0.1, 0.12]},
"mix_kpt": [2, [3, 3, 3], [4, 4, 4, 1, 1, 1]]
}
}
-
example_template: A list of examples, each of which must contain at least aSTRUfile. -
mix_input: Combinations of INPUT parameters. If specified, all combinations of the parameters will be generated for each example. For instance,{"ecutwfc": [50, 60, 70], "kspacing": [0.1, 0.12]}will generate 3*2 = 6 INPUT files. To combine parameters, use '|' to connect them, with values also combined by '|'. For example:{"ecutwfc|kspacing": ["50|0.1", "60|0.12", "70|0.13"]}. -
mix_kpt: Combinations of KPT parameters that automatically generate K points by Gamma-centered Monkhorst-Pack method. If specified, all combinations of the parameters will be generated for each example. There are three ways to define KPT parameters:Thus, an example of
mix_kptcould be:[2, [3, 3, 3], [4, 4, 4, 1, 1, 1]]-
A single integer defines the number of K points in the a/b/c directions, with a shift of (0 0 0) in G space. For example, 4 represents 4 4 4 0 0 0.
-
Three integers define the number of K points in the a/b/c directions, with a shift of (0 0 0) in G space. For example,
[4, 4, 4]represents 4 4 4 0 0 0. -
Three integers define the number of K points in the a/b/c directions, and three floats define the shift in G space. For example,
[4, 4, 4, 1, 1, 1]represents 4 4 4 1 1 1.
-
NOTE: if the mixed inputs are large than 1, then abacustest will create sub-folders named 000000, 000001, 000002, etc. in each example folder.
abacustest supports perturbing configurations to generate new structures for ABACUS calculations.
The param.json may be like:
{
"prepare": {
"example_template": ["example1", "example2"],
"pert_stru": {
"pert_number": 10,
"cell_pert_frac": 0.01,
"atom_pert_dist": 0.1,
"mag_rotate_angle": 5,
"mag_tilt_angle": 10,
"mag_norm_dist": 0.05
}
}
}
-
pert_stru: Random structure perturbations. If specified, perturbed structures will be generated for each example.-
pert_number: The number of perturbed structures. If set to 0, no perturbed structures will be generated. This is the final number of examples, with each new structure perturbed based on the parameters below. -
cell_pert_frac: The maximum perturbation fraction for the unit cell. If null, the cell will not be perturbed. For example, 0.01 means the cell vectors will be perturbed by a maximum of 1%. -
atom_pert_dist: The maximum perturbation distance for atoms, in Angstrom. -
mag_rotate_angle: The rotation angle of magnetic moments for all magnetic atoms, in degrees. If null, magnetic moments will not be rotated. All magnetic atoms will be rotated by the same angle. -
mag_tilt_angle: The tilt angle for magnetic moments, in degrees. If null, magnetic moments will not be tilted. The tilt is applied to all magnetic atoms, but for each atom the tilt angle is randomly chosen. -
mag_norm_dist: The distance of magnetic moments from the normal direction, in$\mu_B$ . If null, magnetic moments will not be displaced.
-
Note 1: Magnetic moment perturbations are only valid for spin-constrained atoms where the "sc" value of at least one magnetic component is 1.
Note 2: Values can also be a list of two numbers, representing the minimum and maximum perturbation values. For example, atom_pert_dist: [0.1, 0.15] means atomic perturbation distances range from 0.1 to 0.15 Angstrom. If both values are the same, the perturbation distance is fixed.
abacustest supports converting ABACUS input files to VASP input files. This is useful for preparing VASP calculations based on existing ABACUS input files.
The param.json may be like:
{
"prepare": {
"abacus2vasp": true,
"potcar": {"H": "H.pot", "O": "O.pot"},
"vasp_setting": {
"ENCUT": 500,
"EDIFF": 1e-5,
"emax_coef": 1.0
}
}
}
-
abacus2vasp: Whether to convert ABACUS inputs to VASP inputs. Defaults to false. Currently supports conversion of cell, coordinates, k-points, standard scf/relax/cell-relax calculations, symmetry, smearing, charge mixing, scf_thr, force_thr, dft_plus_u, nupdown, lspinorb, noncolin, and atomic magnetic settings. EDIFF is set based on scf_thr: scf_thr/1e-2 for PW and scf_thr/1e-1 for LCAO. -
potcar: The path to POTCAR files or a dictionary of POTCARs (e.g.,{"H": "H.psp8", "O": "O.psp8"}). If a path, subfolders named after elements should contain POTCAR files, which the organizational form is consistent with the pseudopotential organizational form provided by VASP (e.g., vasp/PAW_PBE). If a dictionary, keys are element names and values are POTCAR filenames. -
vasp_setting: Additional settings for VASP, and two types of settings:-
Specify values in INCAR. For example:
"ENCUT": 500, "EDIFF": 1e-5, .... -
Special settings:
- "emax_coef": The coefficient for ecutwfc. The actual ENCUT = E_MAX * emax_coef, where E_MAX is the recommended value in the POTCAR. If not defined, ENCUT = ecutwfc * Ry2eV. If ENCUT is also defined, it will be used directly.
-
abacustest supports converting ABACUS input files to QE input files. This is useful for preparing QE calculations based on existing ABACUS input files.
The param.json may be like:
{
"prepare": {
"abacus2qe": true,
"qe_setting": {
"version": "7.0",
"system": {
"ibrav": 0,
"ecutwfc": 50
}
}
}
}
-
abacus2qe: Whether to convert ABACUS inputs to QE inputs. Defaults to false. Currently supports conversion of cell, coordinates, k-points, standard scf/relax/cell-relax calculations, symmetry, smearing, mixing, scf_thr, force_thr, stress_thr, atomic magnetic settings, and hubbard U settings. -
qe_setting: Additional settings for QE inputs. There are three types of settings:-
Specify parameters in "system", "control", "electrons", "ions", "cell". For example:
"system": {"ibrav": 0}, .... -
Specify a block, where the key is the block title and the value is a list of all lines in the block. For example:
"HUBBARD (ortho-atom)": ["U Fe1-3d 5.3", ...]. -
Special keys:
- "version": The QE version, defaulting to 7.0. QE input formats vary between versions.
-
abacustest supports converting ABACUS input files to CP2K input files. This is useful for preparing CP2K calculations based on existing ABACUS input files.
The param.json may be like:
{
"prepare": {
"abacus2cp2k": true,
"cp2k_setting": {
"FORCE_EVAL": {
"DFT": {
"SCF": {
"EPS_SCF": 1e-6
}
}
}
}
}
}
-
abacus2cp2k: Whether to convert ABACUS inputs to CP2K inputs. Defaults to false. Currently supports conversion of cell, coordinates, k-points, calculation type, force_thr, stress_thr, smearing, mixing, and scf_thr (for PW = abacus_value1e3, for LCAO = abacus_value1e2). -
cp2k_setting: Additional settings for CP2K inputs. Should be a dictionary where keys are CP2K input names and values are the corresponding input values. For example:{"FORCE_EVAL": {"DFT": {"SCF": {"EPS_SCF": 1e-6}}}}.
Additional parameters that control file handling and auxiliary settings.
-
link_example_template_extra_files: Bool, Whether to link extra files in each example folder. Defaults to true (links or copies all files in the example folder). If false, only copies or links files required for ABACUS jobs (e.g., INPUT, KPT, STRU, .upf, .orb). -
extra_files: List, Additional files to be linked to each generated example folder. For example:["abc.py", "def.json"]. -
dpks_descriptor: Str, The descriptor for DeepKS calculation. If specified, the dpks file will be linked to each example.
NOTE: All prepare parameters can be used in combination. The actual running sequence of abacustest prepare is as follows:
- Convert the structure defined in
strusto ABACUS input files - Use the generated input files as
example_template, and perform configuration perturbations to generate a series of sub-examples. - Do INPUT/KPT mixing for each sub-examples.
- Convert all generated sub-examples from ABACUS to inputs of other software.
In these processes, the following rules apply:
- When both
example_templateandstrusare specified,example_templateis ignored. - When
example_templatecontains INPUT or KPT, andinput_templateorkpt_templateis specified at the same time, the INPUT/KPT in example_template is ignored.
abacustest supports submitting jobs to Bohrium and other supercomputers.
Before preparing to submit calculations, you need to prepare the examples (each example is a folder that containing the inputs for one ABACUS job), as well as an additional configuration file (such as "param.json") that defines the calculation parameters and settings.
NOTE: since tasks are transferred to remote computing resources (such as Bohrium or other supercomputer platforms), do not use absolute paths in INPUT or STRU files to define pseudopotential or orbital files.
A configuration file mainly contains the following content:
{
"bohrium_group_name": "abacustest",
"save_path":"results",
"max_parallel": 100,
"run_dft":[
{
"ifrun": true,
"sub_save_path": "123",
"image": "registry.dp.tech/dptech/abacus-stable:LTSv3.10",
"example":[["00[0-2]"],"00[3-5]"],
"group_size" : 1,
"bohrium": {
"scass_type":"c8_m16_cpu",
"job_type":"container",
"platform":"ali"
},
"command": "mpirun -np 8 abacus > log",
"extra_files":[]
}
]
}
Where:
save_path: Defines the path to save the results. If this key is not defined, or is set to "" orNone, the value will be replaced by the path defined by "-s" (a parameter ofabacustest; the default for "-s" is "result").max_parallel: Defines the maximum number of parallel jobs. The default is 100.run_dft: Defines details of job execution. The value is a list of dictionaries, and you can set any number of dictionaries.ifrun: If set tofalse, this part will be skipped.sub_save_path: The path withinsave_pathwhere results of this part will be saved, meaning the actual save path will be "save_path/sub_save_path". If this key is not defined, or is set to "" ornull, the actual save path will be "save_path".image: Defines the image name. You can also use names defined inABBREVIATION. The program will first check if the image name is defined inABBREVIATION; if so, the name will be replaced with the value fromABBREVIATION; otherwise, the image name remains as specified.example: The folder names of your jobs. Assuming you have 5 jobs with folder names 000, 001, 002, ..., 004, you can write this as ["000", "001", "002", "003", "004"] or use glob patterns like ["00[0-4]"].group_size: Defines how many example groups to run on a single machine, default is 1.bohrium: If you want to submit jobs to Bohrium, this key must be set. If Bohrium is not used, please remove it.command: The command to run the job, which is the same for all jobs defined inexample.extra_files: If additional files are needed (e.g., "collectdata-abacus.json"), they can be defined here. These files will be copied to each example folder before the command is executed.
The bohrium can be replaced by dispatcher (https://docs.deepmodeling.com/projects/dpdispatcher/en/latest/), which support submitting jobs to more platforms, such as supercomputers. The dispatcher is a dictionary, and the content is like:
"dispatcher": {
"machine_dict":
{
"remote_root": "/home/username/work_path",
"remote_profile": {
"hostname": "xxx.xx.xxx.xxx",
"username": "Username",
"password": "password",
"port": 22
}
},
"resources_dict": {
"number_node": 1,
"cpu_per_node": 8,
"gpu_per_node": 1,
"queue_name": "Normal"
}
}
Before submitting a job, you need to set below environment variables:
export BOHRIUM_USERNAME="your_bohrium_username"
export BOHRIUM_PASSWORD="your_bohrium_password"
export BOHRIUM_PROJECT_ID="your_bohrium_project_id"
These variables are used to authenticate your Bohrium account.
To submit a job, you can use the command:
abacustest submit -p param.json &
where "&" is used to run the command in the background, and abacustest will automatically download the results after the job is finished.
After the job is submitted, you will see output similar to:
Workflow has been submitted (ID: abacustest-kbrb2, UID: f14c5d95-655c-47b4-a709-c9a8138a40cf)
Workflow link: https://workflows.deepmodeling.com/workflows/argo/abacustest-kbrb2
job ID: abacustest-kbrb2, UID: f14c5d95-655c-47b4-a709-c9a8138a40cf
You can track the flow by using your browser to access the URL:
https://workflows.deepmodeling.com/workflows/argo/abacustest-kbrb2?tab=workflow
You can track the job status by visiting the URL provided in the output. If the terminal is closed after the job is submitted, you can download the results later by using the command:
abacustest download -p param.json <ID>
where <ID> is the job ID of the submitted job, such as abacustest-kbrb2 in the above example. Notice that the job ID may be invalid after a certain period of time, then you can download the results by using the UID, such as f14c5d95-655c-47b4-a709-c9a8138a40cf in the above example.
NOTICE: if configuration file defines "prepare" section, then the abacustest will prepare the input files before submitting the job, and in this case, the results of "prepare" will be saved in current directory. If "example" is not defined in "run_dft" section, the generated inputs by "prepare" will be used as "example", while if "example" is defined in "run_dft", jobs in "example" will be used.
abacustest supports extracting key values from the output of ABACUS/QE/VASP jobs. This functionality is valuable for retrieving critical metrics from calculation results, including energy, forces, stress, and other relevant parameters.
You need to prepare a parameter configuration file (e.g., param.json) that specifies the key values to be collected, formatted as follows:
{"PARAM":
["natom","kpt","ibzk","nelec","nbands","force","stress",
"scf_steps","total_time","force_time","stress_time","energy_per_atom","band_gap",
{"SELF_DEFINED_NAME": "An eval string to get the value"}]
}
Only the key "PARAM" is recognized by collectdata, its value is a list of target keys. These keys can be strings or dictionaries:
- If a string, it must correspond to a key predefined in the result class.
- If a dictionary, the key represents a custom name, and the value is an evaluation string enclosed in {} to compute the desired value.
Examples:
- To calculate energy per atom: {"energy per atom": "{energy}/{natom}"}(where "energy" and "natom" are predefined keys, and the result is their quotient).
- To rename the "energy" key to "energy (eV)": {"energy (eV)": "energy"}.
Executing the command:
abacustest collectdata -p param.json -j job1 job2 job3
will collect the specified keys from the output of the jobs job1, job2, and job3. The results will be saved in a JSON file named metrics.json in the current directory.
If you want to save the results in a different file, you can use the -o option.
If you want to collect data for QE/VASP jobs, you can use the -t option to specify the job type, such as -t 0 for ABACUS, -t 1 for QE, and -t 2 for VASP. If not specified, it defaults to ABACUS.
The structure of the metrics.json file is as follows:
{
"job1": {
"natom": 10,
"nelec": 20,
"nbands": 12,
"band_gap": 1.5
},
...
}
Where each key corresponds to a job name and contains the collected metrics.
You can set --outparam to print out all of the keys of one type. Such as: collectdata --outparam 1 -t 0 will print out the keys of ABACUS:
Job type: abacus, all modules: ['abacus']
version: abacus:Abacus.GetVersion() the version of ABACUS
ncore: abacus:Abacus.GetNcore() the mpi cores
normal_end: abacus:Abacus.GetNormalEnd() if the job is nromal ending
INPUT: abacus:Abacus.GetInputParameter() a dict to store the setting in OUT.xxx/INPUT
kpt: abacus:Abacus.GetKptParam() list, the K POINTS setting in KPT file
nbands: abacus:Abacus.GetLogParam() number of bands
converge: abacus:Abacus.GetLogParam() if the SCF is converged
...
The avial modules are listed in the first line, and you can add some of them if needed.
You can also use this collectdata function in your python script by from abacustest.lib_collectdata.collectdata import RESULT.
For example, if you want to collect the total energy, force, and stress from an ABACUS job, you can use the following code:
from abacustest.lib_collectdata.collectdata import RESULT
abacusresult = RESULT(fmt="abacus",path="abacusjob")
# path is the path to the ABACUS job folder, such as "abacusjob"
# fmt is the format of the job, such as "abacus", "qe", "vasp".
energy = abacusresult["total_energy"]
force = abacusresult["force"]
stress = abacusresult["stress"]
If the key is not supported by the RESULT class, the value will be None.