aiida-mlip

Machine learning interatomic potentials AiiDA plugin

status stable

AiiDA >=2.7,<3

General information

Install pip install aiida-mlip

Source code https://github.com/stfc/aiida-mlip

Documentation https://stfc.github.io/aiida-mlip/

Author(s) Federica Zanca, Elliott Kasoar, Jacob Wilkins, Alin M. Elena

Python import import aiida_mlip

Latest version 0.4.1

Released 2025-11-17

Registry checks

All checks passed

Plugins provided

Calculations 5 Parsers 5 Data 2

Entry points

CalcJobs and calculation functions aiida.calculations

mlip.descriptors

Calcjob implementation to calculate MLIP descriptors. Methods ------- define(spec: CalcJobProcessSpec) -> None: Define the process specification, its inputs, outputs and exit codes. prepare_for_submission(folder: Folder) -> CalcInfo: Create the input files for the `CalcJob`.

Input	Required	Valid types	Description
`arch`	false	`Str, NoneType`	Mlip architecture to use for calculation, defaults to mace
`calc_per_atom`	false	`Bool, NoneType`	Calculate descriptors for each atom.
`calc_per_element`	false	`Bool, NoneType`	Calculate mean descriptors for each element.
`code`	false	`AbstractCode, NoneType`	The `Code` to use for this job. This input is required, unless the `remote_folder` input is specified, which means an existing job is being imported and no code will actually be run.
`config`	false	`JanusConfigfile, NoneType`	Name of the log output file
`device`	false	`Str, NoneType`	Device on which to run calculation (cpu, cuda or mps)
`invariants_only`	false	`Bool, NoneType`	Only calculate invariant descriptors.
`log_filename`	false	`Str, NoneType`	Name of the log output file
`metadata`	false
`model`	false	`ModelData, NoneType`	Mlip model used for calculation
`monitors`	false	`Dict`	Add monitoring functions that can inspect output files while the job is running and decide to prematurely terminate the job.
`out`	false	`Str, NoneType`	Name of the xyz output file
`precision`	false	`Str, NoneType`	Precision level for calculation
`remote_folder`	false	`RemoteData, NoneType`	Remote directory containing the results of an already completed calculation job without AiiDA. The inputs should be passed to the `CalcJob` as normal but instead of launching the actual job, the engine will recreate the input files and then proceed straight to the retrieve step where the files of this `RemoteData` will be retrieved as if it had been actually launched through AiiDA. If a parser is defined in the inputs, the results are parsed and attached as output nodes as usual.
`struct`	false	`StructureData, NoneType`	The input structure.

Output	Required	Valid types	Description
`log_output`	true	`SinglefileData`
`remote_folder`	true	`RemoteData`	Input files necessary to run the process will be stored in this folder node.
`results_dict`	true	`Dict`	The `results_dict` output node of the successful calculation.
`retrieved`	true	`FolderData`	Files that are retrieved by the daemon will be stored in this node. By default the stdout and stderr of the scheduler will be added, but one can add more by specifying them in `CalcInfo.retrieve_list`.
`std_output`	true	`SinglefileData`
`xyz_output`	true	`SinglefileData`
`remote_stash`	false	`RemoteStashData`	Contents of the `stash.source_list` option are stored in this remote folder after job completion.

Exit status	Message
1	The process has failed with an unspecified error.
2	The process failed with legacy failure mode.
10	The process returned an invalid output.
11	The process did not register a required output.
100	The process did not have the required `retrieved` output.
110	The job ran out of memory.
120	The job ran out of walltime.
131	The specified account is invalid.
140	The node running the job failed.
150	{message}
160	{message}
305	Some output files missing or cannot be read

mlip.md

Calcjob implementation to run geometry MD calculations using mlips. Methods ------- define(spec: CalcJobProcessSpec) -> None: Define the process specification, its inputs, outputs and exit codes. prepare_for_submission(folder: Folder) -> CalcInfo: Create the input files for the `CalcJob`.

Input	Required	Valid types	Description
`arch`	false	`Str, NoneType`	Mlip architecture to use for calculation, defaults to mace
`code`	false	`AbstractCode, NoneType`	The `Code` to use for this job. This input is required, unless the `remote_folder` input is specified, which means an existing job is being imported and no code will actually be run.
`config`	false	`JanusConfigfile, NoneType`	Name of the log output file
`device`	false	`Str, NoneType`	Device on which to run calculation (cpu, cuda or mps)
`ensemble`	false	`Str, NoneType`	Name for thermodynamic ensemble
`log_filename`	false	`Str, NoneType`	Name of the log output file
`md_kwargs`	false	`Dict, NoneType`	Keywords for molecular dynamics
`metadata`	false
`model`	false	`ModelData, NoneType`	Mlip model used for calculation
`monitors`	false	`Dict`	Add monitoring functions that can inspect output files while the job is running and decide to prematurely terminate the job.
`precision`	false	`Str, NoneType`	Precision level for calculation
`remote_folder`	false	`RemoteData, NoneType`	Remote directory containing the results of an already completed calculation job without AiiDA. The inputs should be passed to the `CalcJob` as normal but instead of launching the actual job, the engine will recreate the input files and then proceed straight to the retrieve step where the files of this `RemoteData` will be retrieved as if it had been actually launched through AiiDA. If a parser is defined in the inputs, the results are parsed and attached as output nodes as usual.
`struct`	false	`StructureData, NoneType`	The input structure.

Output	Required	Valid types	Description
`final_structure`	true	`StructureData`
`log_output`	true	`SinglefileData`
`remote_folder`	true	`RemoteData`	Input files necessary to run the process will be stored in this folder node.
`results_dict`	true	`Dict`	The `results_dict` output node of the successful calculation.
`retrieved`	true	`FolderData`	Files that are retrieved by the daemon will be stored in this node. By default the stdout and stderr of the scheduler will be added, but one can add more by specifying them in `CalcInfo.retrieve_list`.
`stats_file`	true	`SinglefileData`
`std_output`	true	`SinglefileData`
`summary`	true	`SinglefileData`
`traj_file`	true	`SinglefileData`
`traj_output`	true	`TrajectoryData`
`remote_stash`	false	`RemoteStashData`	Contents of the `stash.source_list` option are stored in this remote folder after job completion.

Exit status	Message
1	The process has failed with an unspecified error.
2	The process failed with legacy failure mode.
10	The process returned an invalid output.
11	The process did not register a required output.
100	The process did not have the required `retrieved` output.
110	The job ran out of memory.
120	The job ran out of walltime.
131	The specified account is invalid.
140	The node running the job failed.
150	{message}
160	{message}
305	Some output files missing or cannot be read

mlip.opt

Calcjob implementation to run geometry optimisation calculations using mlips. Methods ------- define(spec: CalcJobProcessSpec) -> None: Define the process specification, its inputs, outputs and exit codes. prepare_for_submission(folder: Folder) -> CalcInfo: Create the input files for the `CalcJob`.

Input	Required	Valid types	Description
`arch`	false	`Str, NoneType`	Mlip architecture to use for calculation, defaults to mace
`code`	false	`AbstractCode, NoneType`	The `Code` to use for this job. This input is required, unless the `remote_folder` input is specified, which means an existing job is being imported and no code will actually be run.
`config`	false	`JanusConfigfile, NoneType`	Name of the log output file
`device`	false	`Str, NoneType`	Device on which to run calculation (cpu, cuda or mps)
`fmax`	false	`Float, NoneType`	Maximum force for convergence
`log_filename`	false	`Str, NoneType`	Name of the log output file
`metadata`	false
`model`	false	`ModelData, NoneType`	Mlip model used for calculation
`monitors`	false	`Dict`	Add monitoring functions that can inspect output files while the job is running and decide to prematurely terminate the job.
`opt_cell_fully`	false	`Bool, NoneType`	Fully optimise the cell vectors, angles, and atomic positions
`opt_cell_lengths`	false	`Bool, NoneType`	Optimise cell vectors, as well as atomic positions
`opt_kwargs`	false	`Dict, NoneType`	Other optimisation keywords
`out`	false	`Str, NoneType`	Name of the xyz output file
`precision`	false	`Str, NoneType`	Precision level for calculation
`properties`	false	`Str, NoneType`	Properties to calculate
`remote_folder`	false	`RemoteData, NoneType`	Remote directory containing the results of an already completed calculation job without AiiDA. The inputs should be passed to the `CalcJob` as normal but instead of launching the actual job, the engine will recreate the input files and then proceed straight to the retrieve step where the files of this `RemoteData` will be retrieved as if it had been actually launched through AiiDA. If a parser is defined in the inputs, the results are parsed and attached as output nodes as usual.
`steps`	false	`Int, NoneType`	Number of optimisation steps
`struct`	false	`StructureData, NoneType`	The input structure.
`traj`	false	`Str, NoneType`	Path to save optimisation frames to

Output	Required	Valid types	Description
`final_structure`	true	`StructureData`
`log_output`	true	`SinglefileData`
`remote_folder`	true	`RemoteData`	Input files necessary to run the process will be stored in this folder node.
`results_dict`	true	`Dict`	The `results_dict` output node of the successful calculation.
`retrieved`	true	`FolderData`	Files that are retrieved by the daemon will be stored in this node. By default the stdout and stderr of the scheduler will be added, but one can add more by specifying them in `CalcInfo.retrieve_list`.
`std_output`	true	`SinglefileData`
`traj_file`	true	`SinglefileData`
`traj_output`	true	`TrajectoryData`
`xyz_output`	true	`SinglefileData`
`remote_stash`	false	`RemoteStashData`	Contents of the `stash.source_list` option are stored in this remote folder after job completion.

Exit status	Message
1	The process has failed with an unspecified error.
2	The process failed with legacy failure mode.
10	The process returned an invalid output.
11	The process did not register a required output.
100	The process did not have the required `retrieved` output.
110	The job ran out of memory.
120	The job ran out of walltime.
131	The specified account is invalid.
140	The node running the job failed.
150	{message}
160	{message}
305	Some output files missing or cannot be read

mlip.sp

Calcjob implementation to run single point calculations using mlips. Attributes ---------- XYZ_OUTPUT : str Default xyz output file name. Methods ------- define(spec: CalcJobProcessSpec) -> None: Define the process specification, its inputs, outputs and exit codes. validate_inputs(value: dict, port_namespace: PortNamespace) -> Optional[str]: Check if the inputs are valid. prepare_for_submission(folder: Folder) -> CalcInfo: Create the input files for the `CalcJob`.

Input	Required	Valid types	Description
`arch`	false	`Str, NoneType`	Mlip architecture to use for calculation, defaults to mace
`code`	false	`AbstractCode, NoneType`	The `Code` to use for this job. This input is required, unless the `remote_folder` input is specified, which means an existing job is being imported and no code will actually be run.
`config`	false	`JanusConfigfile, NoneType`	Name of the log output file
`device`	false	`Str, NoneType`	Device on which to run calculation (cpu, cuda or mps)
`log_filename`	false	`Str, NoneType`	Name of the log output file
`metadata`	false
`model`	false	`ModelData, NoneType`	Mlip model used for calculation
`monitors`	false	`Dict`	Add monitoring functions that can inspect output files while the job is running and decide to prematurely terminate the job.
`out`	false	`Str, NoneType`	Name of the xyz output file
`precision`	false	`Str, NoneType`	Precision level for calculation
`properties`	false	`Str, NoneType`	Properties to calculate
`remote_folder`	false	`RemoteData, NoneType`	Remote directory containing the results of an already completed calculation job without AiiDA. The inputs should be passed to the `CalcJob` as normal but instead of launching the actual job, the engine will recreate the input files and then proceed straight to the retrieve step where the files of this `RemoteData` will be retrieved as if it had been actually launched through AiiDA. If a parser is defined in the inputs, the results are parsed and attached as output nodes as usual.
`struct`	false	`StructureData, NoneType`	The input structure.

Output	Required	Valid types	Description
`log_output`	true	`SinglefileData`
`remote_folder`	true	`RemoteData`	Input files necessary to run the process will be stored in this folder node.
`results_dict`	true	`Dict`	The `results_dict` output node of the successful calculation.
`retrieved`	true	`FolderData`	Files that are retrieved by the daemon will be stored in this node. By default the stdout and stderr of the scheduler will be added, but one can add more by specifying them in `CalcInfo.retrieve_list`.
`std_output`	true	`SinglefileData`
`xyz_output`	true	`SinglefileData`
`remote_stash`	false	`RemoteStashData`	Contents of the `stash.source_list` option are stored in this remote folder after job completion.

Exit status	Message
1	The process has failed with an unspecified error.
2	The process failed with legacy failure mode.
10	The process returned an invalid output.
11	The process did not register a required output.
100	The process did not have the required `retrieved` output.
110	The job ran out of memory.
120	The job ran out of walltime.
131	The specified account is invalid.
140	The node running the job failed.
150	{message}
160	{message}
305	Some output files missing or cannot be read

mlip.train

Calcjob implementation to train mlips. Attributes ---------- DEFAULT_OUTPUT_FILE : str Default stdout file name. Methods ------- define(spec: CalcJobProcessSpec) -> None: Define the process specification, its inputs, outputs and exit codes. validate_inputs(value: dict, port_namespace: PortNamespace) -> Optional[str]: Check if the inputs are valid. prepare_for_submission(folder: Folder) -> CalcInfo: Create the input files for the `CalcJob`.

Input	Required	Valid types	Description
`mlip_config`	true	`JanusConfigfile`	Config file with parameters for training
`code`	false	`AbstractCode, NoneType`	The `Code` to use for this job. This input is required, unless the `remote_folder` input is specified, which means an existing job is being imported and no code will actually be run.
`fine_tune`	false	`Bool, NoneType`	Whether fine-tuning a model
`foundation_model`	false	`ModelData, NoneType`	Model to fine-tune
`metadata`	false
`monitors`	false	`Dict`	Add monitoring functions that can inspect output files while the job is running and decide to prematurely terminate the job.
`remote_folder`	false	`RemoteData, NoneType`	Remote directory containing the results of an already completed calculation job without AiiDA. The inputs should be passed to the `CalcJob` as normal but instead of launching the actual job, the engine will recreate the input files and then proceed straight to the retrieve step where the files of this `RemoteData` will be retrieved as if it had been actually launched through AiiDA. If a parser is defined in the inputs, the results are parsed and attached as output nodes as usual.

Output	Required	Valid types	Description
`checkpoints`	true	`FolderData`
`compiled_model`	true	`SinglefileData`
`logs`	true	`FolderData`
`model`	true	`ModelData`
`remote_folder`	true	`RemoteData`	Input files necessary to run the process will be stored in this folder node.
`results_dict`	true	`Dict`	The `results_dict` output node of the training.
`retrieved`	true	`FolderData`	Files that are retrieved by the daemon will be stored in this node. By default the stdout and stderr of the scheduler will be added, but one can add more by specifying them in `CalcInfo.retrieve_list`.
`remote_stash`	false	`RemoteStashData`	Contents of the `stash.source_list` option are stored in this remote folder after job completion.

Exit status	Message
1	The process has failed with an unspecified error.
2	The process failed with legacy failure mode.
10	The process returned an invalid output.
11	The process did not register a required output.
100	The process did not have the required `retrieved` output.
110	The job ran out of memory.
120	The job ran out of walltime.
131	The specified account is invalid.
140	The node running the job failed.
150	{message}
160	{message}
305	Some output files missing or cannot be read

Data node types aiida.data

mlip.config
aiida_mlip.data.config:JanusConfigfile
mlip.modeldata
aiida_mlip.data.model:ModelData

CalcJob parsers aiida.parsers

mlip.descriptors_parser
aiida_mlip.parsers.descriptors_parser:DescriptorsParser
mlip.md_parser
aiida_mlip.parsers.md_parser:MDParser
mlip.opt_parser
aiida_mlip.parsers.opt_parser:GeomOptParser
mlip.sp_parser
aiida_mlip.parsers.sp_parser:SPParser
mlip.train_parser
aiida_mlip.parsers.train_parser:TrainParser