v1.0.0 Released

DMFF v1.0.0 Released!

New features:

• Implemented QEQ electrostatic potential under constant charge and constant potential modes.
• Refactored the API layer, enhancing frontend robustness.
• Support calculating electrostatic potential in non-periodic systems for point charge, BCC charge, QEQ model, and multipole model.
• Added dmff.api.DMFFTopology class for storing topological information.
• Added dmff.api.ParamSet class, supporting Pytree data structures for improved management of force field parameters.
• Improved computation efficiency of PME kernel on GPU.
• Implemented backend/save_dmff2tf.py, which stores DMFF potential as a TensorFlow module for use in C/C++ API calls.
• Developed an OpenMM plugin, enabling the use of DMFF potential within OpenMM.

v0.2.0 Released

DMFF v0.2.0 Released!

New features:

• Support statistical property optimization using differentiable MBAR estimator
• Support NBFix in Lennard-Jones potential
• Support assigning parameters based on SMIRKS-pattern
• Support bond charge correction (BCC) and virtual site
• Support building neighborlist with freud package

v0.1.1 Released

DMFF v0.1.1 Released!

• Fix bugs in parsing multiple force field files
• Fix bugs in LJ force
• Refine examples and installation code

v0.1.0 Released!

DMFF v0.1.0 released!

• Support differentiable calculation of classical, multiple polarizable and subgragh neural network force fields
• Refined documents and examples
• CI/CD workflows