Update of Cifar10 example which leads to a FID of 3.5

[kilianFatras]

We have improved our training of Flow Matching methods and now we get FID around 3.5. The trick was to use a constant learning rate with exponential moving average and gradient clipping. This has led us to use a smaller network than before (num_channels 128), a smaller batch size than before (128) and a smaller number of iterations (400k). All other findings and details will be found in the update arXiv paper.

Fixes #63

Before submitting

• Did you make sure title is self-explanatory and the description concisely explains the PR?
• Did you make sure your PR does only one thing, instead of bundling different changes together?
• Did you list all the breaking changes introduced by this pull request?
• Did you test your PR locally with pytest command?
• Did you run pre-commit hooks with pre-commit run -a command?

Did you have fun?

Make sure you had fun coding 🙃

[codecov-commenter]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (1c44a6a) 34.23% compared to head (fbc49b6) 34.23%.

❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@           Coverage Diff           @@
##             main      #65   +/-   ##
=======================================
  Coverage   34.23%   34.23%           
=======================================
  Files          55       55           
  Lines        6268     6268           
=======================================
  Hits         2146     2146           
  Misses       4122     4122           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.