One issue that has appeared while I was discovering other bugs is the fact that distributions of optical depths doesn't really match fiducials with the new speed-up updates. I've already discovered some solutions which are approximate, and a better approach might be necessary, that includes different bubble samplers.
One issue that has appeared while I was discovering other bugs is the fact that distributions of optical depths doesn't really match fiducials with the new speed-up updates. I've already discovered some solutions which are approximate, and a better approach might be necessary, that includes different bubble samplers.