Formerly: The Variational Material Integrator
A robust and differentiable simulator for the statics and dynamics of large deformation continuum mechanics, with a focus on designing Neuromechanical Autoencoders.
Uses Automatic Differentiation as a first-class citizen. Optimization is done directly by computing gradients and sparse Hessians of the potential and kinetic energies. Differentiation through the nonlinear solver is carried out through adjoint methods.
For information on usage, check out the Wiki
conda env create -f environment.yml
conda activate varmint
export PYTHONPATH=$(pwd):$PYTHONPATH
Then create a file called local_config.py containing:
exp_root = USE_YOUR_ROOT_EXPERIMENT_DIRECTORY
source_root = ABSOLUTE_ROOT_DIRECTORY_OF_REPO