search

nipreps / eddymotion

Open-source eddy-current and head-motion correction for dMRI.
https://nipreps.org/eddymotion
Apache License 2.0
11 stars 16 forks source link

Should parallelization be handled upstream? #174

Open arokem opened 2 months ago

arokem commented 2 months ago

This is a suggestion to simplify this:

https://github.com/nipreps/eddymotion/blob/b1f70cc67417ac2fab43ccf5196b9b26397b2326/src/eddymotion/model/base.py#L160

We are in the midst of some work to parallelize model fitting in DIPY itself: https://github.com/dipy/dipy/pull/2593, which would benefit all downstream uses of these models, including here. I wonder whether we should aim to use these methods as implemented there (e.g., by passing through parallelization kwargs to DIPY's multi-voxel parallelization methods), instead of implementing that separately here.

oesteban commented 2 months ago

Yes :)

arokem commented 2 months ago

OK - well - let's keep this open if that's fine by you, so that we remember to refactor that code once the parallelization code is incorporated into DIPY (hopefully in the upcoming release).

arokem commented 1 week ago

Good news! https://github.com/dipy/dipy/pull/2593 got merged 😄

arokem commented 1 week ago

Actually, reopening, because we still need to remove the parallelization code from here.