Martin Schubert
Martin Schubert
(I intentionally avoided the use of top/bottom in variable names because of the conflict this creates)
One thing to have in mind is that some "postprocessing" functions (e.g. for computing farfield profiles) rely on sampling the BZ with a regular grid. I suspect that generalizing these...
It would be pretty useful here to have a general `3D -> (stack of 2D slices, slice heights)` function, where the slice heights and values in each slice are optimized...
As a next step, we should collect problematic examples. This could be done in a lightweight way, e.g. with a colab notebook.
This would indeed be useful. I am curious about the issue you are anticipating: would `jax.jacrev` not be adequate? Is it somehow incompatible with mpi4jax?
OK, I did a quick check to see if we have an issue. I believe (fingers crossed) that we are safe, and `jacrev` will just work. ```python from fmmax import...
Thanks @ianwilliamson, yes this makes sense. By the way, we could make use `jacrev` by simply returning the primal as an auxiliary quantity and using `has_aux=True`, as follows: ```python def...
Hi @lucasgrjn , thanks for chiming in and sharing these suggestions. I definitely appreciate discussions and contributions! In general, I see several potential tasks here. The first relates to examples...
Yes, I am also optimistic about some of the new FDTD tools being developed ([luminescent.jl](https://github.com/paulxshen/Luminescent.jl) is another one). One thing that I am keen to retain is ease of install...
Hi @VikramReddy-Anapana , can you paste the code that is causing you issues?