The failing test from the mac env actually happens because your second commit doesn't have all of the changes to pymc.step_methods.hmc.quadpotential that I had over in my PR

So there was a broken intermediate commit?

I guess this is missing: 2913e4a

It's strange though, the test passes locally for me

Ah must be the other mp_ctx that Mac defaults to. Should we be worried though that is not being passed?

I guess this is missing: 2913e4a

Yes, it's that commit.

Ah must be the other mp_ctx that Mac defaults to. Should we be worried though that is not being passed?

It's not super important at the moment I think. The parallel sampling should still work because sampling state is not passed around there, but if you cherry-pick that other commit, then things should pass.

It was the sequential sampling (on the CI we only have 1 core). I had to reorder so we handle the default values before we do the fix on sequential sampling

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 92.85%. Comparing base (5d51953) to head (9dcd50a).
Report is 14 commits behind head on main.

@@            Coverage Diff             @@
##             main    #7637      +/-   ##
==========================================
+ Coverage   92.83%   92.85%   +0.01%     
==========================================
  Files         106      106              
  Lines       17748    17771      +23     
==========================================
+ Hits        16477    16501      +24     
+ Misses       1271     1270       -1