rascaline
rascaline copied to clipboard
Luthaf
Computing representations for atomistic machine learning
From @clecust in #270, split into it's own issue ## GPU Support for rascaline.torch.LodeSphericalExpansion I encountered an issue when attempting to use rascaline.torch.LodeSphericalExpansion() with GPU input data. It appears that...
Following on from PR #237, and completes a TODO item in #239 , the next public API function implemented is the `correlate_tensors` method. This takes two arbitrary TensorMaps and performs...
The first 'iteration' of the Python redesign is starting to converge - see #237. In a rough order of chronology and/or complexity, the next steps to be tackled in future...
It would be nice for new users to have some functionality (a function, a part of the documentation, …) to convert from other packages' hypers to rascaline. I think we...
There are many applications for which it's interesting to be able to assign weights to the atoms - from alchemical weighting models to experimental structures with partial occupations. It would...
Density reconstructions don't work for some choices of hypers. We should make a how-to on density reconstructions and fix all the bugs
On the top of my head, there are the following: - SOAP: - https://doi.org/10.1103/PhysRevMaterials.7.045802 - https://arxiv.org/abs/2310.07604 - LODE: - https://arxiv.org/abs/2308.13208 - I think https://arxiv.org/abs/2304.08966, @andreagrisafi can you confirm? - Hamiltonian...
Include an option to precompute LODE geometric scattering factor (radial projections and spherical harmonics evaluated at k) when the cell is kept fixed in the dataset.
In the process of writing a new API for Clebsch Gordan iterations (see PR #237), it became apparent that generating pseudo-outputs of the CG calculation process (whereby `TensorMaps` with the...
A small tutorial introducing the differences with the normal Python API would be quite helpful. It would need a discussion of forward vs backward mode gradients.