This is major release with several infrastructural changes that will make future development of dynesty easier, as well as significant speedups. The release also deprecated some rarely used options.
- Major refactor of the code that makes it easier to add new samplers.
- Major speedup of the uniform sampler when sampling in parallel
- You can now provide samplers as classes as arguments NestedSampler(sample=RWalkSampler(walks=44))
- Convergence statistics are saved for each iteration that can help diagnose sampling issues
- You can now save the function evaluation history for all function calls throughout the sampling, even when doing it in parallel
- The Hamiltonian slice sampler was removed because of lack of proper testing and not being mantainable
The notebook highlights some of the key changes.