Gregory Desvignes

Thanks for the quick feedback. After some more testing, I found that the main issue was the likelihood, previously computed in single precision only. Then I doubled the number of...