You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
We still support numpy-based functions for building Jacobians and forward modelling in Spline and elsewhere. These are a bit of a pain to maintain and we have to keep testing with and without numba. The numba version is much faster, parallel, and more memory efficient. With the state of numba these days, it's not worth keeping it as an optional dependency.
We should remove the numpy-based functions and add numba as a required dependency. This can allow us to experience with numba functions for other performance critical parts of the project, like BlockReduce.
The text was updated successfully, but these errors were encountered:
We still support numpy-based functions for building Jacobians and forward modelling in
Spline
and elsewhere. These are a bit of a pain to maintain and we have to keep testing with and without numba. The numba version is much faster, parallel, and more memory efficient. With the state of numba these days, it's not worth keeping it as an optional dependency.We should remove the numpy-based functions and add numba as a required dependency. This can allow us to experience with numba functions for other performance critical parts of the project, like
BlockReduce
.The text was updated successfully, but these errors were encountered: