GPUMD-v4.8

GPUMD-v4.8

• Updates for the nep executable:

  • Changed to also fine-tune the descriptor parameters for the fine_tune keyword, which has been tested to lead to higher accuracy than fine-tuning the neural network parameters only. #1295
  • For qNEP models, added lambda_q and lambda_z to set the loss weights for the total charge (presently zero) and Born effective charge (BEC). #1293

• New gnep executable:

  • This has been in existence for quite a few months, but only became mature recently. It provides an alternative way to train certain versions of NEP potential models.

• Updates for the gpumd executable:

  • The fix keyword is in effective for the minimize sd keyword, which can be used to fix atoms during minimization. #1319

GPUMD-v4.7

GPUMD-v4.7

• Updates for the nep executable:

  • Changed to use a new typewise cutoff scheme, in which one can specify a set of radial and angular cutoffs for each species. #1263
  • Changed to use a new ZBL typewise cutoff scheme, in which the inner cutoff is set to 0 instead of half of the outer cutoff, and the default factor is changed from 0.65 to a better value of 0.7. #1265 #1274
    We suggest the following usage for adding the ZBL potential:

      zbl 2.5
      use_typewise_cutoff_zbl 0.7
    

  • Updated a tool to enable the extraction of Born effective charge (BEC) from VASP output. #1272

• Updates for the gpumd executable:

  • Removed the -DUSE_TABLE compliling flag, as this feature does not lead to significant gain of computational performance and has some restrictions. #1258
  • Added the add_spring keyword to support friction simulation. #1273
  • Added the kspace keyword to switch between PPPM and Ewald for qNEP in MD simulation. #1256 #1193

GPUMD-v4.6

New features

• The new features released here actually have been developed during the past several months. They only become mature now. The features are about the new qNEP method, where q stands for charge. In this method, partial charges are assigned to the atoms and the electrostatic (ES) energy for the systems is calculated. The ES energy is combined with the original NEP energy to fit to the target energy. In this way, the partial charges on the atoms can be self-consistently determined. The relevant new and updated keywords are as follows:
  • New keyword charge_mode for the nep executable
  • Updated descriptions for the train.xyz/test.xyz files
  • New keyword compute_dpdt for the gpumd executable
  • Updated dump_xyz keyword for the gpumd executable
  • Updated add_efield keyword for the gpumd executable
    Detailed descriptions of the techniques will be presented in a research paper.

Bug fixes and improvements

• Making the reading of EAM/alloy potential files more robust. #1226

GPUMD-v4.5

New features for the gpumd executable:

• Added the ADP potential. #1185

GPUMD-v4.4

New features for the gpumd executable:

• Added a compute_orientorder keyword to calculate the Steinhardt bond-orientational order parameter. #1134

Bug fixes and improvements for the nep executable:

• Fixed a bug in the training of the temperature-NEP model. #1149
• Extended the save_potential keyword to allow for saving the nep.restart file periodically. #1142

Bug fixes and improvements for the gpumd executable:

• Fixed a bug for the minimize fire keyword when the simulation box is small. #1164
• Fixed a bug in mc vcsgc when only a group of atoms are active. #1167
• Fixed an integer overflow problem for ensemble ti_spring during parameter checking. #1148
• Improved the robustness for ensemble npt_mttk by more stringent input requirements. #1144
• Extended the active keyword to support outputting the per atom uncertainties to active.xyz. #1139

GPUMD-v4.3

GPUMD-v4.3

New features:

• Added the Tersoff-ILP potential model: #1127 #1062

Bug fixes and improvements:

• Improved the minimizer slightly: #1068
• Improved the compute_msd keyword with multiple groups: #1087 #1085
• Fixed a bug for training NEP with multiple GPUs: #1098
• NEP-MD is sped up for the "small-box" case: #1122

GPUMD-v4.2

GPUMD-v4.2

• Improved robustness of ensemble nphug. #1057
• Allowed for using the move keyword in ensemble heat_lan to enable nanoindentation simulation with a fixed layer (fix), a thermostated layer (ensemble heat_lan with zero temperature difference in two groups), a Newtonian layer (nothing), and a moving part (move). #1056
• Allowed for dumping grouping methods using the dump_xyz keyword. #1054
• Allowed for computing the MSD for all the groups in a grouping method in one go. #1051
• Fixed a bug for the compute_extrapolation keyword, which occurs when calculating the extrapolation grade of a system with fewer number of chemical types than that in nep.txt (such as H2O for a H-O-Cl nep potential). #1045
• Fixed a small bug in ensemble ti_liquid. #1041

GPUMD-v4.1

GPUMD-v4.1

This version contains some bug fixes and improvements for GPUMD-v4.0:

• Fixed compiling issue for PLUMED introduced in GPUMD-v4.0. #1008
• Fixed PIMD-NPT (isotropic and orthogonal cases) bugs introduced in GPUMD-v4.0. #1035
• Added the save_potential keyword in nep.in that is useful for saving potential models frequently during fine-tuning of the NEP89 model. #1012 #1031
• Small improvements for ensemble ti_as and bug fix for dump_shock_nemd. #1022

GPUMD-v4.0

GPUMD-v4.0 release

• The nep executable

  • Extended the upper limit of l_max from 4 to 8. #754
  • Enabled outputting per-atom and per-structure descriptors. #832 #837
  • Added the option to write energy_weight in train.xyz. #877
  • Enabled training TNEP models with atomic reference values. #907
  • Published the NEP89 model, a universal NEP model for both inorganic and organic materials across 89 elements. #1000
  • Added the fine_tune keyword in nep.in to allow for finetuning the NEP89 model. #932 #959

• The gpumd executable

  • Added the ILP+SW potential. #761
  • Added the ILP+NEP potential. #835 #860 #951
  • Enabled box optimization for minimize. #841
  • Implemented compute_adf for calculating angular distribution function. #875
  • Added a keyword compute_angular_rdf to calculate angular dependent RDF. #884
  • Added a keyword compute_extrapolation to calculate the extrapolation level of NEP during MD run. #936
  • Added the dump_xyz keyword, which is more general than the dump_exyz keyword. #976
  • Added interface to the DP model. #827 #981
  • Implemented the general EAM/Alloy potential. #994
  • Implemented free energy calculation method for liquids using thermodynamic integration. #999

• Common parts

  • Enabled HIP compiling. #768 #822

• Breaking changes:

  • Changed to always output the 9-component box in thermo.out. #854
  • Changed to output 9 instead of 3 virial components using the compute keyword. #876
  • Changed to calculate the stress tensor using instant velocities instead of target temperature. #946

GPUMD-v3.9.5

• Deprecated:

  • [GPUMD and NEP] NEP2 is not supported any more. Now there are only NEP3 and NEP4. #689 #687

• New or enhanced features:

  • [NEP] Introduced use_typewise_cutoff and use_typewise_cutoff_zbl keywords for NEP training, which could lead to more flexible cutoff schemes and improved performance. #660 #684
  • [GPUMD] Allow for creating multiple xyz files using dump_exyz during a run. #629
  • [GPUMD] Output the total electron stopping power loss during the run with electron_stop. #634
  • [GPUMD] Add electric field (add_efield) based on given fixed charges. #662
  • [GPUMD] Add force (add_force) to a group of atoms. #658
  • [GPUMD] When velocity is missing from run.in, now the code initializes the velocities with a temperature of 300 K, instead of showing the previous "undefined" behavior.
  • [GPUMD] correct_velocity can now be used to correct the velocities for different groups of atoms. #668
  • Better error reporting when reading training/test datasets. #699 #697

• Bug fixes:

  • [GPUMD] For the compute_rdf command, stop calculation when any periodic thickness is smaller than 2.5 times of the RDF cutoff. #649
  • [GPUMD] Fix a bug introduced after adding dipole/polarizability prediction in GPUMD-v3.9.2. The NEP energy was wrongly calculated when multi-GPU is used for MD runs. #661