Skip to content

Latest commit

 

History

History
120 lines (96 loc) · 5.43 KB

UPDATES.rst

File metadata and controls

120 lines (96 loc) · 5.43 KB
Raw

Updates

0.11.7

  • elastic_constants_dynamic calculation completely reworked to use the deformation-fluctuation hybrid method which converges better than the old deformation-only way.
  • point_defect_mobility calculation added that evaluates the point defect mobility energy barrier associated with point defects jumping from one stable position to a nearby position using nudged-elastic band (NEB) calculations.
  • grain_boundary_static calculation added that explores and relaxes a grain boundary configuration. Calculation still in development and will likely result in the other grain boundary calculations being removed.
  • relax_static now has a raise_at_maxcycles that can be set to False allowing calculations that reach maxcycles to finish successfully rather than raising an error.
  • relax_dynamic removed computing the stress/atom as it was unused and doesn't work with some of the newer ML potentials. The master_prepare settings now have separate at_temp and at_temp_50K branches reflecting that the higher temperatures are incrementally prepared from the lower temperature results.
  • The free_energy and free_energy_liquid calculations' master_prepare settings have been added.
  • Base Calculation updated so all calculations now have a URL field, clean(), clean_files() methods, and a calc_output_files list of files generated by the calculation. The URL field provides a means of assigning a persistent identifier (PID) to a calculation's record once uploaded to a CDCS database. The clean_files() method deletes all files listed in calc_output_files allowing for the work space where a calculation was performed to be cleaned up. The clean() method resets the calculation's status to 'not calculated', clears any error message, and deletes the calc_files if requested.
  • The clean_files operation as described above has also been added as a command line option.
  • Fix across multiple calculations to dump at maxiter rather than maxeval so that dump files are created at the end of unconverged minimizations rather than throwing errors.
  • QuickCheck class added that provides a simple means of running many of the cheaper iprPy calculation in succession and collects the results in a clear manner. This is primarily designed to support potential development by allowing a set of target properties to be specified and quickly evaluated as an initial screening step of potential fits.
  • Base IprPyDatabase updated with a finish_bad_calculations() method that finishes calculations in a run directory by setting their status to "error", and uploading both the record and calculation tar to the database. This is largely used to clean up calculations that fail to finish in a reasonable amount of time.
  • The prepare and master_prepare methods are updated for faster and more efficient operations by reducing the database access overhead.
  • runner has been made slightly smarter when selecting a calculation to bid on.
  • A BaseEmperorPrepare class has been added that adds an object-oriented means of defining different calculation pools for a given machine. This further abstracts from master_prepare and gets one step closer to true automated calculation preparation.
  • Updates for newer versions of importlib.resources.
  • Updates form newer versions of Bokeh for iprPy.analysis.PropertyProcessor.
  • Bug fix related to interpreting lammps_command values with the LammpsCommands subset.

0.11.6

  • relax_liquid now has the option to restart and continue incomplete simulations.
  • All records now have a database parameter giving them a default database to pull more content from, if needed. This makes it possible to add methods that retrieve information if it is not directly stored in the records themselves.
  • iprPy retrieve command line updated to make database specification optional and allow both the remote and local databases to be changed.
  • More potentials now included in the old_pots list.

0.11.5

  • phonon calculation: bug fixed, Notebook updated, and analysis methods added.
  • relax_liquid master_prepare options added and updated to reflect how the IPR workflow is being performed.
  • raise_error option added to calculation runs to raise any calculation error rather than save the error message to the results.json file.
  • Records updated to reflect the changes to queries in yabadaba 0.2.0.

0.11.4

  • free_energy calculation method added that evaluates the Gibbs free energy of a solid structure through thermodynamic integration from a Einstein solid.
  • phonon calculation fixed and now properly performs QHA. Default parameter values updated to give better results.
  • relax_dynamic default parameter values for at_temp master prepare changed.
  • Analysis tools for building PropertyProcessor records added to iprPy from previously private code.
  • The StackingFaultMap2D calculation/record class now reads in and interprets results from the JSON/XML files.
  • XSL transformations added for some calculations.
  • Bug fix for crystal_space_group calculation's pandasfilter method.
  • Bug fix for E_vs_r_scan calculation's cdcsquery method.
  • Fix in diatom_scan calculation for potentials that don't like multiple atypes.
  • Bug fix with running surface_energy_static and point_defect_static calculations, where "E_coh" is now properly updated to "E_pot".