[14:14:10] o/ [14:14:13] I'm back! [14:14:20] And feeling much better than yesterday [14:14:39] I basically slept from noon --> 7AM [14:37:27] \o_ [15:13:24] Good morning guillom & J-Mo :) [15:13:34] Hello hello [15:13:50] hey guillom and halfak! [15:13:53] halfak: welcome back. [15:14:28] o/ JohanJ [15:15:04] JohanJ, did you see progress on https://phabricator.wikimedia.org/T131450? [15:17:27] Now I did! (I got stuck on stuff related to the Wikimedia Sweden court case today, so haven't really had time to check Phabricator.) Will take a look at what Josve05a has done. [15:17:55] Great! Happy to be making progress on svwiki :) [15:18:10] We're going to be standing up a lot of new models due to this hackathon [15:19:10] That's the great thing about hackathons. :) [15:19:27] Indeed. :) [15:19:44] Lots of new volunteers. mwahaha. [15:27:57] halfak: I love that "privacy policy" (integritetspolicy) is one of the Swedish false positives. [15:31:17] lol yeah. We find a lot of usernames in there too. [15:51:48] hey YuviPanda. I've been getting the 'too many redirects' error when I try to start my paws server. Happened yesterday, and today. Any idea what's up? [16:19:25] halfak, I set up a meeting for us at 3pm central tomorrow to talk Teahouse. You can move it if that time doesn't work for you (for example, if you think your jet lag may render you unconscious during the late afternoon) [16:22:02] hey J-Mo [16:22:18] I'm going to spend about 10mins trying to find an underlying cause and if that fails just manually fix it just for you [16:22:31] thx [16:26:58] J-Mo: try again now [16:27:09] I dug in and found the bug, filed at https://github.com/yuvipanda/jupyterhub-kubernetes-spawner/issues/5 [16:28:07] cool. thanks, YuviPanda. Any workarounds in the meantime? Is this a random, intermittent bug, or something I can avoid by altering my behavior? [16:28:26] J-Mo: random intermittent, unfortunately. [16:28:56] are you able to log in to your paws server consistently? Because this bug strikes me every time (at least since yesterday) [16:29:37] J-Mo: I am yeah. but I also only login once and then my pods usually stay on, so you're more likely to hit it than I am [16:30:11] J-Mo: however, I think this should be an easily fixable bug. I'll try to get it fixed this week - we're doing a kubernetes upgrade today that should help [16:30:12] huh, well shut me up! Right after complaining, now it let's me on :) [16:30:45] J-Mo: :) yeah, one of the annoying bits about this bug is that once it hits you it stops new servers from spawning. [16:30:59] for you, at least! so I manually killed your old server [16:31:17] RIP My Old Server [16:31:19] so it does spawn servers, except somehow forgets it has done so and says 'nope, I have not' and ends up stuck in a loop [16:31:31] oh, reminds me: YuviPanda, am I currently able to publish notebooks publicly? [16:32:21] J-Mo: vaguely/kinda/nope. [16:32:28] another thing I hope to fix this week [16:32:42] J-Mo: there's paws-public.wmflabs.org but that's not really usable. [16:33:13] cool. I'll stop perstering you and let you get back to work then :) Thanks dude! [16:33:57] J-Mo: np! I was at the jupyter core devs meeting all of last week (I've merge rights now!), so didn't get much done [17:46:04] o/ halfak [17:46:20] o/ sabya [17:47:19] any chance you have documented hashingvectorizer modelling you talked about? [17:47:29] Oh yeah! Sorry to be so slow with that. [17:48:18] Was hoping to give you some code that made a simple example of model train/test. I've never actually implemented a hashingvectorizer yet [17:51:12] I'm interested in hearing out about modeling problem :) [17:52:42] OK. Will be rushing to get some patches ready for a breaking change we're deploying on Thursday 4/7 [17:53:14] I'll likely get you example around 1700 on Friday 4/8 [17:53:38] I might try to build up a simple hashing vectorizer in the example :) [17:54:07] that'll be great :) [18:01:48] someday I'll be smart enough (or rather have enough time!) to understand all of these :D [18:02:46] YuviPanda, you might like this example since it will show ML like it is an API that you simply program against! [18:03:08] interesting! [18:03:36] I can understand the words but don't know enough about ML to have an intuitive feeling for it (yet!) [18:08:50] soon :) [18:10:42] yeah. I think once I get k8s and PAWS to a stable state (and we've hired another labs ops :D) I'll have a bit more time [18:14:26] Would be great to have your help with some performance optimization in ORES. We just merged a feature extraction profiling system [18:14:33] nice. [18:14:55] Mixing some optimization and feature selection, we should be able to speed up the models quite a lot. [18:15:04] Doing this right will involve modeling work too. [18:15:20] I suppose your extra free time is far out so we might beat you to it. [18:15:35] But then again, there's lots of things to do and I'm just this guy, you know? [18:16:54] https://en.wikipedia.org/wiki/Zaphod_Beeblebrox#Achievements [18:17:00] If the reference was lost [18:17:07] "Vell, Zaphod's just zis guy, you know?" [18:19:44] hehe :D [18:19:45] * sabya feels ecstatic, seeing a reference to one of my fav sci-fi novel [18:20:04] * YuviPanda was referencing the Total Perspective Vortex yesterday [18:20:19] :) [18:40:47] time for bed [18:40:50] o/ [19:54:43] Are business cards still a thing in academic conferences these days? I've never really had a use for them and always end up with hundreds of unused cards, so I haven't ordered any since my job changed 2 years ago. [19:55:41] guillom: I ended up with a bunch and am now not sure what all those people said. I did have very intersting conversations tho [19:57:27] I feel like I should just tell the one or two people whom I'd give a card to to google my name. [19:58:25] (The difficulty is: They probably couldn't spell it.) [19:58:39] guillom, physical reminders to send an email [19:58:43] I use 'em for that [19:59:00] I try to makea note on the card about something I'll remember for a person and then take a picture of it [19:59:03] If someone doesn't have a business card, I'll just have them write their email address/name down on one of mine or in my notebook. [19:59:20] +1 chasemp [19:59:25] But do I really want people to send me e-mails? hmmm :) Getting their email address and emailing them feels better. [19:59:33] When I'm being best halfak, I write a note on the cards. [19:59:44] guillom, fair point [20:00:06] I guess the heart of the problem is unrelated: I don't like to talk to strangers. [20:00:08] :p [20:00:16] I have a few cards from maybe 8 years ago and each tiem I think I'll throw them out a little voice says, "..maybe you'll still email" [20:00:22] this is like hoarder brain [20:00:35] Same here [20:00:58] I have a pile of people that I sort of should email, but I don't want to take on a new conversation. [20:01:55] I think I'll stick to not having cards printed for now. Worst case I'll take a stack of post-it notes and use that. [20:02:20] Thank you for the rubberducking! [20:02:25] cards tends to be proxies for a linked in request nowadays :) [20:02:31] heh [20:03:03] * guillom never uses Linkedin either except for that one time each month to accept invitations. [20:03:35] (As you may have guessed, I'm terrible at this "professional networking" thing.) [20:05:34] guillom, or awesome! I mean, look at all the cool people you have in your immediate network :) [20:08:03] guillom: are we LinkedIn®????? [20:08:29] Emufarmers: You added me a few months ago IIRC. [20:08:40] halfak: heh. [20:09:10] gw team