Two of my boxes have started getting a bunch of validate errors on Separation 1.46 (both CPU and GPU) work units. Can someone look at them and let me know what could be causing the errors.
Thanks for any help.

---

Validation matches returned results to a certain degree of precision; unfortunately, in some cases different hardware can result in slightly different calculation results (depending [mostly?] on rounding methods) and there are certain data sets for which these discrepancies get bigger and bigger as the results approach optimization.

If you look at the most recent News item New Separation Runs 6/9/2021, Tom Donlon introduced the newest set of runs thus:

I've just put some new separation runs up on the server. Remember those stripe 84 and 85 runs that would start to throw validate errors as they became more optimized? I've been testing and comparing runs on different builds and *hopefully* that problem has been resolved.

And (as Tom points out) this is not a new issue -- it seems to be inherent in the data!

Hope this helps...

Cheers - Al.

Hi Al,
Thanks for the info. I guess there is nothing I can do on my end to stop this.
Should I post this in Tom's thread too?

---

Hi Al,
Thanks for the info. I guess there is nothing I can do on my end to stop this.
Should I post this in Tom's thread too?

Looks like he's picked up on this, so I suspect he'll ask for more evidence if he needs it...

Cheers - Al.

He's going to try and try again releasing these Stripe 84 and 85 tasks because they are the last bit of data he needs to finish his paper.
He thought he fixed the problem with those datasets but they have reoccurred again. Needs to take another crack at it.

---