Add is_sync logic to Metric

[tchaton]

Before submitting

• Was this discussed/approved via a Github issue? (no need for typos and docs improvements)
• Did you read the contributor guideline, Pull Request section?
• Did you make sure to update the docs?
• Did you write any new necessary tests?

What does this PR do?

This PR makes synchornization, state_dict and load_dict generation fully manual for the users, so they have full control on their metric states.

for i in range(steps):

    metric(i)

    metric.sync()
    state_dict = metric.state_dict() # synced states
    assert metric.is_synced 

    metric.unsync()
    state_dict = metric.state_dict() # local states
    assert not metric.is_synced 

Fixes # (issue).

PR review

Anyone in the community is free to review the PR once the tests have passed.
If we didn't discuss your PR in Github issues there's a high chance it will not be merged.

Did you have fun?

Make sure you had fun coding 🙃

[pep8speaks]

Hello @tchaton! Thanks for updating this PR.

There are currently no PEP 8 issues detected in this Pull Request. Cheers! 🍻

Comment last updated at 2021-07-01 17:43:39 UTC

[codecov]

Codecov Report

Merging #339 (5d33441) into master (18056a5) will decrease coverage by 20.40%.
The diff coverage is 45.23%.

@@             Coverage Diff             @@
##           master     #339       +/-   ##
===========================================
- Coverage   96.37%   75.97%   -20.41%     
===========================================
  Files         111      112        +1     
  Lines        3586     3600       +14     
===========================================
- Hits         3456     2735      -721     
- Misses        130      865      +735     

Flag	Coverage Δ
Linux	75.97% <45.23%> (-0.02%)	⬇️
Windows	75.97% <45.23%> (-0.02%)	⬇️
cpu	75.97% <45.23%> (-20.35%)	⬇️
gpu	?
macOS	75.97% <45.23%> (-20.35%)	⬇️
pytest	75.97% <45.23%> (-20.41%)	⬇️
python3.6	?
python3.8	?
torch1.3.1	?
torch1.4.0	?
torch1.9.0	?

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

Impacted Files	Coverage Δ
torchmetrics/metric.py	57.18% <43.90%> (-38.36%)	⬇️
torchmetrics/utilities/exceptions.py	100.00% <100.00%> (ø)
torchmetrics/utilities/distributed.py	17.24% <0.00%> (-81.04%)	⬇️
torchmetrics/image/kid.py	18.39% <0.00%> (-78.17%)	⬇️
torchmetrics/image/inception.py	30.00% <0.00%> (-65.00%)	⬇️
torchmetrics/image/fid.py	32.55% <0.00%> (-51.17%)	⬇️
torchmetrics/classification/auc.py	50.00% <0.00%> (-50.00%)	⬇️
torchmetrics/classification/kldivergence.py	37.50% <0.00%> (-50.00%)	⬇️
torchmetrics/functional/classification/auroc.py	44.92% <0.00%> (-40.58%)	⬇️
torchmetrics/functional/classification/accuracy.py	55.71% <0.00%> (-38.58%)	⬇️
... and 70 more

---

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 18056a5...5d33441. Read the comment docs.

[justusschock]

Looks good so far, some comments :)

[chauhang]

@tchaton Model saving after DDP training is working fine with this fix. Tested for single and multi-gpus.