Closed sethaxen closed 1 month ago
Remaining errors seem to be ones introduced in #304 and unrelated to this PR
Remaining errors seem to be ones introduced in https://github.com/TuringLang/Bijectors.jl/pull/304 and unrelated to this PR
Uhmm that's strange o.O Don't understand why this wasn't failing in the original PR. Ooor it might be because it hit the cholesky error and thus didn't run the interface tests on 1.6.. Should be a quick Compat.jl inclusion though; lemme have a check
It seems #314 does not address the issue on Julia 1.6.
I don't think #314 was completed before merging. Its CI was still failing with a similar error.
Yes, #314 was indeed not ready for a merge @yebai ; why was it merged?
Yes, https://github.com/TuringLang/Bijectors.jl/pull/314 was indeed not ready for a merge @yebai ; why was it merged?
my bad -- it shouldn't have been merged.
@sethaxen I just pushed the fix directly to this branch. Let's see if CI succeeds now:)
Damn, even this doesn't work because eachslice
only supports a single value as the dims
arg :confused:
Really sorry about this @sethaxen ; this bug was hidden behind an unrelated numerical issue that caused this particular test to never be run on 1.6. But the cause of this shouldn't have been merged.
I'll just disable those tests on this PR and then we'll have to fix it in a separate PR.
Seems like it worked!
EDIT: and, no problem, @torfjelde !
Lovely! Feel free to hit the big green button:)
Sadly, I am not an "authorized user," and the button is gray.
Want me to do it then? Also happy to let give you authorization given your involvement in Bijectors.jl if you want to:)
Want me to do it then? Also happy to let give you authorization given your involvement in Bijectors.jl if you want to:)
Sure to both!
Done:) Wonderful stuff @sethaxen ; thanks!
This PR implements the numerical suggestions in #301. It does not make any of the suggested renaming changes, which will be left for a future PR.