[16:04:11] hello everyone [16:04:29] everybody [16:04:36] everybody or everyone? [16:18:04] dsaez: o/ [16:18:06] welcome! [16:18:13] rather welcome back! [16:18:17] thanks! [16:18:36] bmansurov, leila, are we having the isdsdfsasd meeting today? [16:18:58] *isdimile joba :D [16:19:06] yes. but I have a conflict and can't attend. Jonathan and Isaac will talk. [16:19:11] dsaez: ^ [16:19:23] oooook [16:19:27] ;) [16:27:08] dsaez: we strike you with a meeting on your first day back 😁 [16:27:49] bmansurov: yes, and I'm also trying to clean my inbox, amazing the amount of mails that you can gather in just 4 days. [16:30:11] 'm' key is your friend [16:30:19] use it without hesitation [16:55:18] dsaez: miriam_: (not sure who else was interested in DC gpus) Any chance you could help make a decision regarding if 8GB is acceptable GPU memory, or if we need to go to 16GB? Does 8GB limit your ability to work? [17:00:04] ebernhadson, in this case bigger is always better ... [17:00:42] 16GB is about double the price and about 10% slower, but if the memory is most important we can go that way [17:01:06] compared to the time waiting by paid employees i'm sure the price isn't that important :) [17:01:11] ebernhardson hi! I don't think we *need* 16GB, it would be better, but 8 GB is acceptable, we will just use smaller batch size. [17:15:53] smaller batch=slightly lower accuracy (negligeble) and longer training time. But I think it's fine [17:19:27] thanks! I've noted that while 16GB would be useful, it's not a blocker. The decision then mostly goes to how much we want to spend on a card that has the potential to never work [17:19:43] (optimistic it will work, but we don't know yet :) [17:24:40] thanks ebernhardson! let's be optimistic :) [19:44:06] hey @leila @isaacj just wanted to check before implementing https://phabricator.wikimedia.org/T213847 [19:45:05] I'm planning to allow surveys to define a list of country codes. The survey will only show up in one of those country codes match window.Geo['country']. Does that help unblock the survey you want to run? [19:47:46] https://www.mediawiki.org/wiki/Extension:QuickSurveys#Defining_an_audience [19:50:02] jdlrobson: thanks yeah, that makes sense. so long as an undefined list means all countries are included [19:58:59] isaacj: I'm in a midterm planning meeting. pull me in for jdlrobson's question as needed. otherwise, I see that you're responding. [20:11:56] thanks jdlrobson too because i see the edit count logic now excludes logged-out users if you set a min-edit threshold! [21:00:29] correct, omitting it will target all countries [21:01:11] bmansurov thanks for https://phabricator.wikimedia.org/T217576 I'll keep you posted if it looks like things change. And I'll leave the start/end times for you to decide. Sounds good? [21:01:34] isaacj: correct. Is the distinction between anons and logged in users targeting thus needed? Setting minEdits: 0, will only target logged in users, but if you want to only target anons, there's no way of doing that right now. [21:02:31] j-mo: ok, sounds good. [21:05:51] jdlrobson: oh yeah, I didn't think of that regarding surveying just anonymous users. we would still need the functionality for surveying only logged-out users as there are use cases where we don't want to interfere with editors or accidentally log information about them [21:06:35] add a maxEdits? [21:35:17] maxEdits: 0 will target editors with edit count 0 as well as anons [21:35:30] isaacj please take https://gerrit.wikimedia.org/r/#/c/mediawiki/extensions/QuickSurveys/+/494544 for a spin and let me know if that works for your suvey! [21:43:59] awesome jdlrobson - will take a look and see what I can do! [21:53:36] not sure there's a real-world difference between the two [21:53:48] maybe an editor with a home on a different wiki [23:10:24] Platonides: there is one crucial difference: 1 of those users created a user account despite not making any edits. The other user didn't create an account but may or may not be editing.