Add ifft helpers, basic tests #94
Conversation
| import pytorch_finufft | ||
|
|
||
|
|
||
| def check_t2_ifft_undoes_t1(N: int, dim: int, device: str) -> None: |
There was a problem hiding this comment.
it strikes me that maybe we need to name t1_ifft to be the one that is actually type 2 ...
There was a problem hiding this comment.
Yeah, happy to discuss naming. We could also be clearer in the docstring which function this inverts, not just which one it wraps
|
|
||
| def check_t1_ifft_undoes_t2(N: int, dim: int, device: str) -> None: | ||
| """ | ||
| Tests that nuifft_type1 undoes nufft_type2 |
There was a problem hiding this comment.
similar Q: should we name the ifft by what type it actually is or what type it is undoing?
WardBrian
force-pushed
the
feature/ifft-helpers
branch
from
0256e83 to
a6cac23
Compare
|
I've made the suggested changes up to the point of a possible re-naming. I did have one question @eickenberg, which is that the finufft docs seem to go out of their way to state that they do not provide inverse transforms. See, e,g: https://finufft.readthedocs.io/en/latest/trouble.html#mathematical-issues-and-advice
Are we using 'inverse' in a different sense than they mean? |
|
Yep, treating the adjoint like the inverse would be bad. Closing this and the related issue |
Closes #87