Closed segasai closed 3 years ago
Changes Missing Coverage | Covered Lines | Changed/Added Lines | % | ||
---|---|---|---|---|---|
py/dynesty/utils.py | 1 | 2 | 50.0% | ||
py/dynesty/sampler.py | 13 | 15 | 86.67% | ||
<!-- | Total: | 55 | 58 | 94.83% | --> |
Files with Coverage Reduction | New Missed Lines | % | ||
---|---|---|---|---|
py/dynesty/utils.py | 1 | 85.74% | ||
<!-- | Total: | 1 | --> |
Totals | |
---|---|
Change from base Build 1188386185: | 0.2% |
Covered Lines: | 3394 |
Relevant Lines: | 4150 |
I have also tentatively replaced the saved_run in the internal sampler to be truncated saved_run from the parent one which is I think the correct thing to do.
And finally I now coded the newly created sampler for sample_batch. There is still some amount of hacks needed, filling saved_run and setting up bounds, but I think it's still clearer and more explicit. and the outputs during the run should be more sensible, I'd believe. Without that change I just didn't feel I was confident enough in correctness of the code.
Great! Let me know when I should take a look at merging this in.
I think it is ready and the 100 tests worked fine too. It is possible that I missed something that the tests don't cover, but that is always a possibility
On 2 Sep 2021 17:37, Josh Speagle @.***> wrote: This email was sent to you by someone outside the University. You should only click on links or attachments if you are certain that the email is genuine and the content is safe.
Great! Let me know when I should take a look at merging this in.
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://github.com/joshspeagle/dynesty/pull/326#issuecomment-911865558, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AAA2XSPFWKZZKGODJYJXAD3T76RZVANCNFSM5DHZQSPA.
The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336. Is e buidheann carthannais a th’ ann an Oilthigh Dhùn Èideann, clà raichte an Alba, à ireamh clà raidh SC005336.
Now out of 100 runs x70 tests there is only one failure 1 FAILED tests/test_dyn.py::test_dyn Which I can't easily fix without just bumping the threshold. I'll probably keep it as it is. I hope that concludes everything that needed fixing before 1.2
get rid of dlog_batch internal variable put in the dlogz=0.01 function parameter (currently not available from the outside yet) Also prevent spurious warnings in the case we are sampling the 'top' part of the posterior. In that case I set the upper logl limit to be +inf. And if that's the case I'm not checking that logl is above loglmax. addresses #325