Releases: rjdirisio/pyvibdmc
Debugging Patch
- Allowed the g matrix to be a function in fixed-node DMC.
- Allowed restarts with importance sampling
Fixed Node Patch
Additional functionality for fixed node DMC.
Fixed Node DMC
Add fixed node capabilities.
Restart bug fixes
Allow for continuous weights to be added in for restarting DMC sims, add in parent continuous weights to wave function output for continuous weight simulations.
Bug fixes and quality of life improvements
Continuous weighting weights are now returned with wave function, harmonic sampling can have varied mass.
Adiabatic DMC, MPI improvements
Incorporation of simple option to do adiabatic DMC, and speed improvements for MPI4Py based on using chunksize
instead of splittiing up the data by myself.
Imp Samp Improvements
Continuous weighting fix for imp sampling and efficiency increase in discrete weighting imp samp.
Atomic Unit Standardization of PyVibDMC
Previously, PyVibDMC
file output was in atomic units, but any analysis or reading of that data was done in angstroms or cm-1. Now, the output has remained in atomic units, and the analysis code is by default still in atomic units. One can pass in an optional keyword argument in order to convert the output to angstroms/cm-1.
NOTE: This is not a backwards compatible change. Please be careful of loading in old data or using old code, now the default is atomic units for everything.
Full Changelog: 1.2.3...1.3.0
Chain Rule Efficiency
Made Chain rule more efficient. Based on how it is currently structured, one can now reuse derivatives that have already been calculated.
Feature: Importance Sampling
In this release, the use of guiding functions is incorporated into the DMC simulation. The guiding functions have multiprocessing and MPI support and use a similar interface as the potential manager. See updated documentation.