MAINT removed useless apply function

[MuawizChaudhary]

Back in the good old days, before I existed, Kymatio reimplemented various calls that a nn.module had. These included .cpu,cuda, and were often implemented using a _apply call.

A reviewer of our JMLR submission kindly noted that we could just inherit from a nn.module instead of reimplementing all those calls.

This _apply_filters tucked and hidden away in the 3d utils code is the last remaining vestige of this old way of doing things.

[codecov-commenter]

Codecov Report

Merging #830 (4bcb6fa) into dev (ca9073f) will increase coverage by 0.10%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##              dev     #830      +/-   ##
==========================================
+ Coverage   88.03%   88.14%   +0.10%     
==========================================
  Files          64       64              
  Lines        2323     2319       -4     
==========================================
- Hits         2045     2044       -1     
+ Misses        278      275       -3     

Flag	Coverage Δ
jenkins_main	88.14% <ø> (+0.10%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
kymatio/scattering3d/utils.py	96.55% <ø> (+8.67%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update ca9073f...4bcb6fa. Read the comment docs.

[lostanlen]

Can't use 'tar -xzf' extract archive file: /home/runner/work/_actions/_temp_735093b6-e20a-47f8-9c61-589d4c653a4c/fb47e774-6995-4eb3-8ca0-770e1e7792dc.tar.gz. return code: 2.

This build-pip CI looks very brittle

[edouardoyallon]

Be extra careful. The reason it was done this way is because we wanted an access to the filters via psi[j] and that the filters were stored as a parameter. Unfortunatly, when using .cuda(), the pointer was not going through the correct direction.

[MuawizChaudhary]

Be extra careful. The reason it was done this way is because we wanted an access to the filters via psi[j] and that the filters were stored as a parameter. Unfortunatly, when using .cuda(), the pointer was not going through the correct direction.

hm, wasnt this because we weren't inheriting nn.module?

A git grep for "_apply_filters" shows that this function is not used anywhere.

[edouardoyallon]

no clue, i'm just saying it was there for this reason

[MuawizChaudhary]

Can we merge?