When converting an ASE-object to a torch-sim SimState object via io.atoms_to_state() the cell is transposed to follow the TorchSim convention. However, when reading in a TorchSimTrajectory and converting it to an ASE-object via the TorchSimTrajectory.get_atoms() function, the cell is not transposed back to ASE convention. A simple fix would be to modify the associated part in the TorchSimTrajectory.get_atoms()- function from cell=np.ascontiguousarray(arrays["cell"])[0] to cell=np.ascontiguousarray(arrays["cell"])[0].T.
When converting an ASE-object to a torch-sim
SimStateobject viaio.atoms_to_state()the cell is transposed to follow the TorchSim convention. However, when reading in aTorchSimTrajectoryand converting it to an ASE-object via theTorchSimTrajectory.get_atoms()function, the cell is not transposed back to ASE convention. A simple fix would be to modify the associated part in theTorchSimTrajectory.get_atoms()- function fromcell=np.ascontiguousarray(arrays["cell"])[0]tocell=np.ascontiguousarray(arrays["cell"])[0].T.