old model was not compatible with CastTo64 transform anymore
new model is trained on rmd17 instead of md17
now implemented:
SpkCalculator
SpkEnsembleCalculator <- inherits from SpkCalculator
BatchwiseCalculator <- inherits from SpkCalculator
BatchwiseEnsembleCalculator <- inherits from SpkEnsembleCalculator
SimpleEnsembleAverage (example for a particular average strategy)
the calculators have two new methods (to allow for more modular classes):
_collect_results <-- calculates model outputs and converts property units
_ase_specific_output_format <-- for individual structures this converts energy output to scalar value, for batches of structures it keeps the output in array form
replaced ethanol model:
now implemented: SpkCalculator SpkEnsembleCalculator <- inherits from SpkCalculator BatchwiseCalculator <- inherits from SpkCalculator BatchwiseEnsembleCalculator <- inherits from SpkEnsembleCalculator SimpleEnsembleAverage (example for a particular average strategy)
the calculators have two new methods (to allow for more modular classes): _collect_results <-- calculates model outputs and converts property units _ase_specific_output_format <-- for individual structures this converts energy output to scalar value, for batches of structures it keeps the output in array form
updated batchwise optimization howto jupyter notebook --> 01_howto_batchwise_relaxations.ipynb
added howto_ensemble_calculations jupyter notebook --> 02_howto_ensemble_calculator