torch-dftd
torch-dftd copied to clipboard
pfnet-research
pytorch implementation of dftd2 & dftd3
This PR adds paramaters for https://pubs.aip.org/aip/jcp/article/148/6/064104/196461/B97-3c-A-revised-low-cost-variant-of-the-B97-D. Also see CP2K implementation: https://github.com/cp2k/cp2k/blob/7123f6cb9512cd8f56af81252de383ca40d906e7/src/qs_dispersion_pairpot.F#L877-L881
To fix the following error > UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before...
Hello All, while running the code I got this error: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single...
In test_torch_dftd3_calculator.py/_assert_energy_force_stress_equal, there is a code below. ``` if np.all(atoms.pbc == np.array([True, True, True])): s1 = atoms.get_stress() s2 = atoms.get_stress() assert np.allclose(s1, s2, atol=1e-5, rtol=1e-5) ``` This code cannot compare...
https://github.com/pfnet-research/torch-dftd/pull/3#discussion_r664202835 random `cell` assignment may be not robust.
For my test case, this implementation was slightly faster than the pymatgen neighborlist but I did not extensively test this. ```python from matscipy.neighbours import neighbour_list def calc_neighbor_by_matscipy( pos: Tensor, cell:...
Hi developers, I have implemented a simple version of lammps-like skin layer for neighbourlist in here: https://github.com/CheukHinHoJerry/torch-dftd (see Readme for a short summary). The motivation of having this is to...
The Grimme group has released an updated and re-parameterized version of their dispersion correction, [DFT-D4](https://github.com/dftd4/dftd4). Would it be possible to add this new parameterization to this torch implementation?
As recommended in [pep 0561](https://peps.python.org/pep-0561/), a blank `py.typed` marker should be included when type hints are used so downstream codes can type check with `mypy` and similar tools.
