@Canakiwi2@YVRCovidPlots@COVIDPoops19 Yeah, they pass the sniff test for me, if you will; here's a quick comparison of the new and old values for each sample. The data already using the new test didn't change. The retested values are all about 5x higher than their original values with the old test.
@mendel I suspect it doesn't matter; my experience on a single-user instance is that the promise of federation is real and only UX difference is the local timeline, which is probably uninteresting on large public instances anyway. Plus moving isn't too painful; followers will follow you.
@seaotta this might happen if the site isn't serving intermediate certificates along with the leaf certificate; caching behaviour can make it a little hard to rationalize
@doachs Thanks for the lead! This is with Ramnode; as far as I know they have their own metal, but I opened a ticket with them to see if they know anything.
I think Google is blackholing my VPS's ipv6 address; I can't use the Google Cloud SDK unless I prefer ipv4 name resolution, & can't curl https://t.co/TW9czdRAQ8. ipv6 seems to be working otherwise on that node. I can't get Google eyes on this without a service contract, can I?
@maskupmelb@YVRCovidPlots@COVIDPoops19 Dunno if this is helpful for you but the first few cells of https://t.co/TpaHxjpxrV demonstrate how to pull the data off the Metro Vancouver website.
@FreeWheelinRob@j_mcelroy Her top 3 were McKechnie ES, Van Horne ES, and Champlain Heights CC; on 21-22% of ballots. Bottom 3 were Britannia CSC, Mount Pleasant CC, Grandview ES; on 5-7.5% of ballots.
@MidniteLibrary @j_mcelroy The polls with the highest fraction of votes allocated to losing candidates were Grandview Elementary (76.5%), Britannia CSC (76.0%), and Carnegie CC (74.5%).
Highest fraction to winning candidates were Kerrisdale CC (63.0%), Shaughnessy ES (61.2%), Trinity Church (60.8%)
@daka_x I don't see a convincing upward trend in any of these plots. Last week's numbers looked maybe elevated but those have come back down. Lions Gate could be the exception but those measurements look quite noisy.
It looks like @HelloPaperspace Core is a compelling alternative that would have saved me some driver downloads and faffing around in complicated control panels
Exercise for this weekend was finding the cheapest Windows cloud node that could run Tensorflow, for a CSU prof friend to throw his students at. Ended up on a Google n1-highmem-2 with a T4 GPU; expecting the all-in cost to be around $450/mo. How'd I do?
it's fun to be called into a statistical consult to deliver the news that nobody needs to run any more statistics 😄 some days you just want a qualitative binary signal and the data oblige
This was the price breakdown; Windows licenses are rather expensive! and pricing depends on the number of cores, which is why I opted for the highmem shape
@yang_yi_cn@YVRCovidPlots@COVIDPoops19 The data are flagged as "No sample collected" but I haven't seen an indication about whether they intend to resume collection or not