This may also address #242

Codecov Report

Attention: Patch coverage is 96.66667% with 2 lines in your changes missing coverage. Please review.

Project coverage is 91.91%. Comparing base (bf26a57) to head (ed082ec).
Report is 33 commits behind head on dev.

@@            Coverage Diff             @@
##              dev     #306      +/-   ##
==========================================
+ Coverage   91.19%   91.91%   +0.72%     
==========================================
  Files          55       55              
  Lines        2306     2277      -29     
==========================================
- Hits         2103     2093      -10     
+ Misses        203      184      -19     

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

NFW and tNFW not passing the test for potential. Any idea why that is?

Ah, that was because I made the new thin lens default reduced deflection angle compute the gradient of the potential (rather than cyclic definition with physical deflection angle). The way that calculation is done involves a vmap, and NFW/TNFW cant do both vmap and autograd. Nothing changed about NFW/TNFW, they still work, just the test now uses vmap and autograd rather than a very large autograd call.

Posting here the improvements to TNFW. Now works with autodif and vmap (NFW now has this too). Also, better handling of F function, is arccosh for 0<x<1-epsilon and is arccos for 1+epsilon<x, for x=1+-epsilon we use taylor expansion to get correct first and second derivatives. in the images below we compare several approximations where F_arccosh is the adopted approximation.