Fix for incorrect usage of detach(), cpu(), to()

[dvolgyes]

What does this PR do?

If you use .detach()/to()/cpu(), you need to assign it to a new variable,
it does not change the source variable.

Fixes #6214

Before submitting

• Was this discussed/approved via a GitHub issue? (not for typos and docs)
• Did you read the contributor guideline, Pull Request section?
• Did you make sure your PR does only one thing, instead of bundling different changes together?
• Did you make sure to update the documentation with your changes? (if necessary)
• Did you write any new necessary tests? (not for typos and docs)
• Did you verify new and existing tests pass locally with your changes?
• Did you update the CHANGELOG? (not for typos, docs, test updates, or internal minor changes/refactorings)

PR review

Anyone in the community is free to review the PR once the tests have passed.
Before you start reviewing make sure you have read Review guidelines. In short, see the following bullet-list:

• Is this pull request ready for review? (if not, please submit in draft mode)
• Check that all items from Before submitting are resolved
• Make sure the title is self-explanatory and the description concisely explains the PR
• Add labels and milestones (and optionally projects) to the PR so it can be classified

Did you have fun?

Make sure you had fun coding 🙃

[codecov]

Codecov Report

Merging #6216 (30258e2) into master (925f082) will decrease coverage by 2%.
The diff coverage is 50%.

@@           Coverage Diff           @@
##           master   #6216    +/-   ##
=======================================
- Coverage      93%     91%    -2%     
=======================================
  Files         159     159            
  Lines       11360   11251   -109     
=======================================
- Hits        10603   10257   -346     
- Misses        757     994   +237     

[tchaton]

Great catch !

[tchaton]

Dear @dvolgyes,

The Result object implements in-place operation. Therefore, your changes won't work.
Therefore, I returned self from to, and cpu functions.

Best,
T.C

[Borda]

@dvolgyes I have applied this to the remaining codebase, mind checkingng it...

[dvolgyes]

I see, thank you for the fix.
In place: I would say it is a very confusing behavior, it looks very pytorch-ish,
but using a different approach (in place).
And in this case, it might not matter that much.
(Some objects i assumed to be tensor might have been lightning objects.)

[dvolgyes]

I looked through, and it looks good to me.
As a remark, the models (nn.Module) are in-place,
so if you call a .cuda() on them, they return themselves,
but there it is not necessary.
I think it is a good practice, because you don't have to keep it in mind
which one works in-place (model, some PL), and which one doesn't (tensors).
So I think the model = model.cpu() is not strictly necessary, but I support the change.

[dvolgyes]

Codecov: I don't understand how could it be that low. Something got accidentally disabled? (Jumping from 80 percent to 20%)

[Borda]

Codecov: I don't understand how could it be that low. Something got accidentally disabled? (Jumping from 80 percent to 20%)

it just takes time till all tests are finished stat aggregated, so the faster is base pkg testing and full testing takes about 15-20min

[dvolgyes]

I see, thanks for the explanation.
I am not familiar with the setup / internals, and I still don't see how this aggregation works,
but I can live without the details. :) Thanks again.