[bugfix] Resolve logging reduction when using sync_dist + reduce_fx={mean, max}

[tchaton]

What does this PR do?

See title.

Fixes #9100
Fixes #4395

Does your PR introduce any breaking changes? If yes, please list them.

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 list all the breaking changes introduced by this pull request?
• Did you update the CHANGELOG? (not for typos, docs, test updates, or internal minor changes/refactorings)

PR review

Anyone in the community is welcome to review the PR.
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 🙃

[ohayonguy]

I am still observing a difference between the logged loss and the loss returned by training_step:

Is this expected? nt_xent_loss should be exactly the same as loss.
When logging a loss in training step, should we even pass sync_dist=True? Or do we only care about doing that for validation steps (for accurate model checkpointing)?

This is my training_step:

[tchaton]

Hey @ohayonguy,

Lightning is adding a smoothing on the loss by computing a running mean over the logged loss within the progress bar.
Personally, I find it quite confusing.
If you feel the same, please on an issue and ask for it to be removed :)

Best,
T.C

[carmocca]

If you feel the same, please on an issue and ask for it to be removed :)

You can comment here: #9372