OOM for large problem sizes

[garrettwrong]

This fixes the OOM condition I recently encountered with full size experiments. I found a place where I was not batching during the evaluation from the stacked Coef to Image. I previously just skirted by, but no longer do with the recently added GPU calls prior to this area.

This adds a batch_size to the related classes and implements a batched evaluation loop.

I'll need to review it more. Got sidetracked with the annoying logger/warnings issue.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 90.63%. Comparing base (ec6a5d7) to head (18a58a9).
Report is 1 commits behind head on develop.

Additional details and impacted files

@@           Coverage Diff            @@
##           develop    #1226   +/-   ##
========================================
  Coverage    90.63%   90.63%           
========================================
  Files          132      132           
  Lines        13698    13705    +7     
========================================
+ Hits         12415    12422    +7     
  Misses        1283     1283           

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

[j-c-c]

LGTM!

[garrettwrong]

Going to merge this to move things along.