MAINT redefine cdgmm3d to call cdgmm in torch skcuda backend

[MuawizChaudhary]

This pull request redefines cdgmm3d to call cdgmm in torch skcuda backend.

I'm also using this as a proxy for testing the torch_skcuda 3d backend.

[codecov-commenter]

Codecov Report

Merging #865 (442a5fb) into dev (4a6d9ea) will increase coverage by 0.18%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##              dev     #865      +/-   ##
==========================================
+ Coverage   87.83%   88.02%   +0.18%     
==========================================
  Files          64       64              
  Lines        2253     2221      -32     
==========================================
- Hits         1979     1955      -24     
+ Misses        274      266       -8     

Flag	Coverage Δ
jenkins_main	88.02% <100.00%> (+0.18%)	⬆️

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

Impacted Files	Coverage Δ
...matio/scattering3d/backend/torch_skcuda_backend.py	100.00% <100.00%> (+19.04%)	⬆️

---

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 4a6d9ea...442a5fb. Read the comment docs.

[lostanlen]

is this tied to a specific issue? what's the plan here?

[MuawizChaudhary]

my reasoning: why are we repeating code that we already have?

[lostanlen]

oh wow, we have duplicate code here? then yes sounds like an easy win

[lostanlen]

is the change from a RuntimeError to a TypeError harmless here?

[MuawizChaudhary]

define harmless?

I changed it because it seemed like we forgot to switch the 3d cgmm over to throwing a Type Error when we changed the 1d and 2d code to be unified

[janden]

my reasoning: why are we repeating code that we already have?

Maybe there was a reason for having a separate cdgmm3d? On the other hand, the tests pass…

[MuawizChaudhary]

my reasoning: why are we repeating code that we already have?

Maybe there was a reason for having a separate cdgmm3d? On the other hand, the tests pass…

The code is the same except for the error checking.

Only one test had to be changed in order for the tests to be passing.

[janden]

Ok sounds good then.