DeePMD-kit’s documentation

Getting Started

• Getting Started
  • 1. Easy install
    • 1.1. Install off-line packages
    • 1.2. Install with conda
    • 1.3. Install with docker
    • 1.4. Install Python interface with pip
  • 2. DeePMD-kit Quick Start Tutorial
    • 2.1. Task
    • 2.2. Background
    • 2.3. Practice

Advanced

• 1. Installation
  • 1.1. Easy install
    • 1.1.1. Install off-line packages
    • 1.1.2. Install with conda
    • 1.1.3. Install with docker
    • 1.1.4. Install Python interface with pip
  • 1.2. Install from source code
    • 1.2.1. Install the python interface
    • 1.2.2. Install the C++ interface
  • 1.3. Install from pre-compiled C library
    • 1.3.1. Use Pre-compiled C Library to build the LAMMPS plugin and GROMACS patch
  • 1.4. Install LAMMPS
    • 1.4.1. Install LAMMPS’s DeePMD-kit module (built-in mode)
    • 1.4.2. Install LAMMPS (plugin mode)
  • 1.5. Install i-PI
  • 1.6. Install GROMACS with DeepMD
    • 1.6.1. Patch source code of GROMACS
    • 1.6.2. Compile GROMACS with deepmd-kit
  • 1.7. Building conda packages
  • 1.8. Install Node.js interface
    • 1.8.1. Install from npm
    • 1.8.2. Build from source
  • 1.9. Easy install the latest development version
    • 1.9.1. Install with docker
    • 1.9.2. Install with pip
    • 1.9.3. Download pre-compiled C Library
    • 1.9.4. Pre-release conda-forge packages
• 2. Data
  • 2.1. System
  • 2.2. Formats of a system
    • 2.2.1. NumPy format
    • 2.2.2. HDF5 format
    • 2.2.3. Raw format and data conversion
  • 2.3. Prepare data with dpdata
• 3. Model
  • 3.1. Overall
    • 3.1.1. Theory
    • 3.1.2. Instructions
  • 3.2. Descriptor "se_e2_a"
    • 3.2.1. Theory
    • 3.2.2. Instructions
  • 3.3. Descriptor "se_e2_r"
    • 3.3.1. Theory
    • 3.3.2. Instructions
  • 3.4. Descriptor "se_e3"
    • 3.4.1. Theory
    • 3.4.2. Instructions
  • 3.5. Descriptor "se_atten"
    • 3.5.1. DPA-1: Pretraining of Attention-based Deep Potential Model for Molecular Simulation
    • 3.5.2. Theory
    • 3.5.3. Introduction to new features of DPA-1
    • 3.5.4. Data format
    • 3.5.5. Training example
  • 3.6. Descriptor "hybrid"
    • 3.6.1. Theory
    • 3.6.2. Instructions
  • 3.7. Determine sel
  • 3.8. Fit energy
    • 3.8.1. Theory
    • 3.8.2. The fitting network
    • 3.8.3. Loss
  • 3.9. Fit spin energy
    • 3.9.1. Spin
    • 3.9.2. Spin Loss
  • 3.10. Fit tensor like Dipole and Polarizability
    • 3.10.1. Theory
    • 3.10.2. The fitting Network
    • 3.10.3. Loss
    • 3.10.4. Training Data Preparation
    • 3.10.5. Train the Model
  • 3.11. Fit electronic density of states (DOS)
    • 3.11.1. The fitting Network
    • 3.11.2. Loss
    • 3.11.3. Training Data Preparation
    • 3.11.4. Train the Model
    • 3.11.5. Test the Model
  • 3.12. Type embedding approach
    • 3.12.1. Theory
    • 3.12.2. Instructions
  • 3.13. Descriptor "se_a_mask"
  • 3.14. Deep potential long-range (DPLR)
    • 3.14.1. Theory
    • 3.14.2. Train a deep Wannier model for Wannier centroids
    • 3.14.3. Train the DPLR model
    • 3.14.4. Molecular dynamics simulation with DPLR
  • 3.15. Deep Potential - Range Correction (DPRc)
    • 3.15.1. Theory
    • 3.15.2. Training data
    • 3.15.3. Training the DPRc model
    • 3.15.4. Run MD simulations
    • 3.15.5. Pairwise DPRc
  • 3.16. Linear model
  • 3.17. Interpolation or combination with a pairwise potential
    • 3.17.1. Theory
    • 3.17.2. Table file
    • 3.17.3. Interpolation with a short-range pairwise potential
    • 3.17.4. Combination with a pairwise potential
• 4. Training
  • 4.1. Train a model
  • 4.2. Advanced options
    • 4.2.1. Learning rate
    • 4.2.2. Training parameters
    • 4.2.3. Options and environment variables
    • 4.2.4. Adjust sel of a frozen model
  • 4.3. Training Parameters
  • 4.4. Parallel training
    • 4.4.1. Tuning learning rate
    • 4.4.2. Scaling test
    • 4.4.3. How to use
    • 4.4.4. Logging
  • 4.5. Multi-task training
    • 4.5.1. Theory
    • 4.5.2. Perform the multi-task training
    • 4.5.3. Initialization from pretrained multi-task model
    • 4.5.4. Share layers among energy fitting networks
  • 4.6. TensorBoard Usage
    • 4.6.1. Highlighted features
    • 4.6.2. How to use Tensorboard with DeePMD-kit
    • 4.6.3. Examples
    • 4.6.4. Attention
  • 4.7. Known limitations of using GPUs
  • 4.8. Finetune the pretrained model
• 5. Freeze and Compress
  • 5.1. Freeze a model
  • 5.2. Compress a model
    • 5.2.1. Theory
    • 5.2.2. Instructions
• 6. Test
  • 6.1. Test a model
  • 6.2. Calculate Model Deviation
    • 6.2.1. Theory
    • 6.2.2. Instructions
    • 6.2.3. Relative model deviation
• 7. Inference
  • 7.1. Python interface
    • 7.1.1. External neighbor list algorithm
  • 7.2. C/C++ interface
    • 7.2.1. C++ interface
    • 7.2.2. C interface
    • 7.2.3. Header-only C++ library interface (recommended)
  • 7.3. Node.js interface
• 8. Command line interface
  • 8.1. Named Arguments
  • 8.2. Valid subcommands
  • 8.3. Sub-commands
    • 8.3.1. transfer
    • 8.3.2. train
    • 8.3.3. freeze
    • 8.3.4. test
    • 8.3.5. compress
    • 8.3.6. doc-train-input
    • 8.3.7. model-devi
    • 8.3.8. convert-from
    • 8.3.9. neighbor-stat
    • 8.3.10. train-nvnmd
    • 8.3.11. gui
• 9. Integrate with third-party packages
  • 9.1. Use deep potential with ASE
  • 9.2. Run MD with LAMMPS
    • 9.2.1. units
    • 9.2.2. Enable DeePMD-kit plugin (plugin mode)
    • 9.2.3. pair_style deepmd
    • 9.2.4. Compute tensorial properties
    • 9.2.5. Long-range interaction
    • 9.2.6. Use of the centroid/stress/atom to get the full 3x3 “atomic-virial”
    • 9.2.7. Computation of heat flux
  • 9.3. Run path-integral MD with i-PI
  • 9.4. Running MD with GROMACS
    • 9.4.1. DP/MM Simulation
    • 9.4.2. All-atom DP Simulation
  • 9.5. Interfaces out of DeePMD-kit
    • 9.5.1. dpdata
    • 9.5.2. OpenMM plugin for DeePMD-kit
    • 9.5.3. Amber interface to DeePMD-kit
    • 9.5.4. CP2K interface to DeePMD-kit
    • 9.5.5. DP-GEN
    • 9.5.6. MLatom
    • 9.5.7. ABACUS

Developer Guide

• Python API
  • backend package
    • Submodules
    • backend.dp_backend module
    • backend.dynamic_metadata module
      • dynamic_metadata()
    • backend.find_tensorflow module
      • find_tensorflow()
      • get_tf_requirement()
      • get_tf_version()
    • backend.read_env module
      • get_argument_from_env()
      • set_scikit_build_env()
  • deepmd package
    • DeepEval
      • DeepEval.build_neighbor_list()
      • DeepEval.eval_typeebd()
      • DeepEval.load_prefix
      • DeepEval.make_natoms_vec()
      • DeepEval.model_type
      • DeepEval.model_version
      • DeepEval.reverse_map()
      • DeepEval.sess
      • DeepEval.sort_input()
    • DeepPotential()
    • DipoleChargeModifier
      • DipoleChargeModifier.build_fv_graph()
      • DipoleChargeModifier.eval()
      • DipoleChargeModifier.modify_data()
    • Subpackages
      • deepmd.cluster package
        • get_resource()
        • Submodules
        • deepmd.cluster.local module
        • deepmd.cluster.slurm module
      • deepmd.descriptor package
        • Descriptor
        • DescrptHybrid
        • DescrptLocFrame
        • DescrptSeA
        • DescrptSeAEbd
        • DescrptSeAEbdV2
        • DescrptSeAEf
        • DescrptSeAEfLower
        • DescrptSeAMask
        • DescrptSeAtten
        • DescrptSeAttenV2
        • DescrptSeR
        • DescrptSeT
        • Submodules
        • deepmd.descriptor.descriptor module
        • deepmd.descriptor.hybrid module
        • deepmd.descriptor.loc_frame module
        • deepmd.descriptor.se module
        • deepmd.descriptor.se_a module
        • deepmd.descriptor.se_a_ebd module
        • deepmd.descriptor.se_a_ebd_v2 module
        • deepmd.descriptor.se_a_ef module
        • deepmd.descriptor.se_a_mask module
        • deepmd.descriptor.se_atten module
        • deepmd.descriptor.se_atten_v2 module
        • deepmd.descriptor.se_r module
        • deepmd.descriptor.se_t module
      • deepmd.entrypoints package
        • compress()
        • convert()
        • doc_train_input()
        • freeze()
        • make_model_devi()
        • neighbor_stat()
        • start_dpgui()
        • test()
        • train_dp()
        • transfer()
        • Submodules
        • deepmd.entrypoints.compress module
        • deepmd.entrypoints.convert module
        • deepmd.entrypoints.doc module
        • deepmd.entrypoints.freeze module
        • deepmd.entrypoints.gui module
        • deepmd.entrypoints.ipi module
        • deepmd.entrypoints.main module
        • deepmd.entrypoints.neighbor_stat module
        • deepmd.entrypoints.test module
        • deepmd.entrypoints.train module
        • deepmd.entrypoints.transfer module
      • deepmd.fit package
        • DOSFitting
        • DipoleFittingSeA
        • EnerFitting
        • Fitting
        • GlobalPolarFittingSeA
        • PolarFittingSeA
        • Submodules
        • deepmd.fit.dipole module
        • deepmd.fit.dos module
        • deepmd.fit.ener module
        • deepmd.fit.fitting module
        • deepmd.fit.polar module
      • deepmd.infer package
        • DeepDOS
        • DeepDipole
        • DeepEval
        • DeepGlobalPolar
        • DeepPolar
        • DeepPot
        • DeepPotential()
        • DeepWFC
        • DipoleChargeModifier
        • EwaldRecp
        • calc_model_devi()
        • Submodules
        • deepmd.infer.data_modifier module
        • deepmd.infer.deep_dipole module
        • deepmd.infer.deep_dos module
        • deepmd.infer.deep_eval module
        • deepmd.infer.deep_polar module
        • deepmd.infer.deep_pot module
        • deepmd.infer.deep_tensor module
        • deepmd.infer.deep_wfc module
        • deepmd.infer.ewald_recp module
        • deepmd.infer.model_devi module
      • deepmd.loggers package
        • set_log_handles()
        • Submodules
        • deepmd.loggers.loggers module
      • deepmd.loss package
        • DOSLoss
        • EnerDipoleLoss
        • EnerSpinLoss
        • EnerStdLoss
        • TensorLoss
        • Submodules
        • deepmd.loss.dos module
        • deepmd.loss.ener module
        • deepmd.loss.loss module
        • deepmd.loss.tensor module
      • deepmd.model package
        • DOSModel
        • DipoleModel
        • EnerModel
        • GlobalPolarModel
        • MultiModel
        • PolarModel
        • WFCModel
        • Submodules
        • deepmd.model.dos module
        • deepmd.model.ener module
        • deepmd.model.frozen module
        • deepmd.model.linear module
        • deepmd.model.model module
        • deepmd.model.model_stat module
        • deepmd.model.multi module
        • deepmd.model.pairtab module
        • deepmd.model.pairwise_dprc module
        • deepmd.model.tensor module
      • deepmd.nvnmd package
        • Subpackages
      • deepmd.op package
        • import_ops()
      • deepmd.train package
        • Submodules
        • deepmd.train.run_options module
        • deepmd.train.trainer module
      • deepmd.utils package
        • DeepmdData
        • DeepmdDataSystem
        • LearningRateExp
        • PairTab
        • Plugin
        • PluginVariant
        • Submodules
        • deepmd.utils.argcheck module
        • deepmd.utils.batch_size module
        • deepmd.utils.compat module
        • deepmd.utils.compress module
        • deepmd.utils.convert module
        • deepmd.utils.data module
        • deepmd.utils.data_system module
        • deepmd.utils.errors module
        • deepmd.utils.finetune module
        • deepmd.utils.graph module
        • deepmd.utils.learning_rate module
        • deepmd.utils.multi_init module
        • deepmd.utils.neighbor_stat module
        • deepmd.utils.network module
        • deepmd.utils.nlist module
        • deepmd.utils.pair_tab module
        • deepmd.utils.parallel_op module
        • deepmd.utils.path module
        • deepmd.utils.plugin module
        • deepmd.utils.random module
        • deepmd.utils.region module
        • deepmd.utils.sess module
        • deepmd.utils.spin module
        • deepmd.utils.tabulate module
        • deepmd.utils.type_embed module
        • deepmd.utils.weight_avg module
    • Submodules
    • deepmd.calculator module
      • DP
        • DP.calculate()
        • DP.implemented_properties
        • DP.name
    • deepmd.common module
      • add_data_requirement()
      • cast_precision()
      • clear_session()
      • expand_sys_str()
      • gelu()
      • gelu_tf()
      • get_activation_func()
      • get_np_precision()
      • get_precision()
      • j_loader()
      • j_must_have()
      • make_default_mesh()
      • safe_cast_tensor()
      • select_idx_map()
    • deepmd.env module
      • GLOBAL_ENER_FLOAT_PRECISION
      • GLOBAL_NP_FLOAT_PRECISION
      • global_cvt_2_ener_float()
      • global_cvt_2_tf_float()
      • reset_default_tf_session_config()
    • deepmd.lmp module
      • get_env()
      • get_library_path()
      • get_op_dir()
  • deepmd_utils package
    • Subpackages
      • deepmd_utils.entrypoints package
        • Submodules
        • deepmd_utils.entrypoints.doc module
        • deepmd_utils.entrypoints.gui module
      • deepmd_utils.loggers package
        • set_log_handles()
        • Submodules
        • deepmd_utils.loggers.loggers module
      • deepmd_utils.model_format package
        • DescrptSeA
        • EmbeddingNet
        • EnvMat
        • FittingNet
        • FittingOutputDef
        • ModelOutputDef
        • NativeLayer
        • NativeNet
        • NativeOP
        • NetworkCollection
        • OutputVariableDef
        • fitting_check_output()
        • get_deriv_name()
        • get_reduce_name()
        • load_dp_model()
        • make_embedding_network()
        • make_fitting_network()
        • make_multilayer_network()
        • model_check_output()
        • save_dp_model()
        • traverse_model_dict()
        • Submodules
        • deepmd_utils.model_format.common module
        • deepmd_utils.model_format.env_mat module
        • deepmd_utils.model_format.network module
        • deepmd_utils.model_format.output_def module
        • deepmd_utils.model_format.se_e2_a module
      • deepmd_utils.utils package
        • Submodules
        • deepmd_utils.utils.argcheck module
        • deepmd_utils.utils.argcheck_nvnmd module
        • deepmd_utils.utils.batch_size module
        • deepmd_utils.utils.compat module
        • deepmd_utils.utils.data module
        • deepmd_utils.utils.data_system module
        • deepmd_utils.utils.errors module
        • deepmd_utils.utils.model_stat module
        • deepmd_utils.utils.pair_tab module
        • deepmd_utils.utils.path module
        • deepmd_utils.utils.plugin module
        • deepmd_utils.utils.random module
        • deepmd_utils.utils.weight_avg module
    • Submodules
    • deepmd_utils.common module
      • add_data_requirement()
      • expand_sys_str()
      • get_np_precision()
      • j_loader()
      • j_must_have()
      • make_default_mesh()
      • select_idx_map()
    • deepmd_utils.env module
      • GLOBAL_ENER_FLOAT_PRECISION
      • GLOBAL_NP_FLOAT_PRECISION
    • deepmd_utils.main module
      • RawTextArgumentDefaultsHelpFormatter
      • get_ll()
      • main()
      • main_parser()
      • parse_args()

• Index

• Module Index

• Search Page