Possible improvements to neighbor lists about jax-md HOT 3 CLOSED

I took a pass at some API improvements in #154. However, some other things (changing the capacity multiplier and the semantics for dr_threshold) didn't make it in yet. I do think they are good suggestions and I hope to add them in next week. At some point (hopefully soon-ish) there will also be another round of neighbor list improvements that target better compatibility with periodic_general and npt simulations.

For reference, the idea behind scaling dr_threshold was that max(sigma) sort of sets the units for the problem. However, I also agree that there are use cases where that is not true and probably keeping dr_threshold a scale free constant might be better in practice.

from jax-md.

Okay! Not quite a week, but as of version 0.1.22 dr_threshold no longer scales with max(sigma) and the neighbor list arguments have been threaded through all of the appropriate energy functions.

Let me know if you run into any issues! Thanks for the feedback.

from jax-md.

Awesome! I'll let you know if we run into any issues :)

from jax-md.