CAN Attention: standard management of headsand out_channels

[gbg141]

Standardizing the CAN attention mechanism. Now, as in other attention-based models, the argument out_channels actually represents the final number of output channels --i.e., it must be divisible by heads to successfully perform the attention mechanism.

Previously the actual output channels were computed as heads * out_channels, which could be misleading.

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 97.37%. Comparing base (2267768) to head (3652f91).

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #276   +/-   ##
=======================================
  Coverage   97.37%   97.37%           
=======================================
  Files          58       58           
  Lines        2054     2055    +1     
=======================================
+ Hits         2000     2001    +1     
  Misses         54       54           

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

[ninamiolane]

YES! Great catch, this could have been very confusing to users.