Codecov Report

❌ Patch coverage is 90.90909% with 8 lines in your changes missing coverage. Please review.
✅ Project coverage is 81.63%. Comparing base (c67edf7) to head (e94fb60).
⚠️ Report is 466 commits behind head on master.

@@            Coverage Diff             @@
##           master    #1041      +/-   ##
==========================================
+ Coverage   81.54%   81.63%   +0.09%     
==========================================
  Files         140      141       +1     
  Lines       15400    15509     +109     
==========================================
+ Hits        12558    12661     +103     
- Misses       2842     2848       +6     

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

At this point, this seems to work. However, sampling single-atom Cartesian displacements, even for alanine dipeptide in vacuum, is SLOW. I think I need to improve the premise of the example before it is useful pedagogically without the context of this PR and #1040.

Performance is also not great. I haven't profiled it yet, but I wonder if we're going to and from the GPU with each MC trial. It looks to me like the OpenMMTools MCMC stuff was designed with hybrid Monte Carlo in mind, where one proposal creation would involve a significant chunk of work on the GPU.

Right now, I can't get an equilibrated trajectory because (1) trajectories are very long because single atom displacements require many frames for a transition; and (2) each frame seems to be quite slow compared to MD. (And of course (3) I need my computer for other things!)