DM-39178: Add deterministic seed to ks-test model sampling. #191
Conversation
| # Seed this with the mean value, so that the same data will get the | ||
| # same result. | ||
| randomSeed = int((mu1 + mu2)/2.) |
There was a problem hiding this comment.
It's an interesting thought. But is it really what you want? Upstream changes which cause changes in the inputs will cause a seed-change - why not just hard code it like we do everywhere else? I mean, I think it's fine, just kinda curious if you really do want to do that...
There was a problem hiding this comment.
I don't know if I fully believe in completely hard-coding seeds outside tests; in fact we often use exp number or something that changes but is deterministic. I just wanted something quick that would give the same output for the same input.
There was a problem hiding this comment.
Ah, I'd forgotten that it was tests I was thinking of. OK, nice, thanks.
There was a problem hiding this comment.
The fact the fit is so on the hairy edge that seeds cause it to fail is a different worry though, but for now, fixing random Jenkins failures is clearly more important.
There was a problem hiding this comment.
Well, the hairy-edge part is going to be fixed on DM-39167.
There was a problem hiding this comment.
OK, I figured, thanks 🙂
No description provided.