[15:05:11] halfak: first translations came today, several languages added or changed . [15:05:31] Cool! [15:05:37] I also thought we should have a train of deployment for wikilabels and ores [15:05:38] I just about to kill our translations epic card [15:05:45] like the mediawiki [15:05:54] * halfak does not really understand deployment trains [15:06:06] I've heard the term before, but I don't know what they are [15:06:10] brb [15:06:13] I explain in more depth [15:06:38] kk [15:11:12] halfak: deployment in mediawiki is like this: once every patch in core or any of registered extensions (including ORES) is merged [15:11:16] it goes directly to beta [15:11:28] and people can see if it breaks anything [15:12:10] but depends on when it's merged it goes to package of mediawiki called mediawiki 1.24.wmf17 [15:13:21] and twice a week the new package got deployed in prod cluster [15:13:24] Amir1, would be nice if we could go to the labs instance as our beta. [15:14:21] except SWAT deployment that are non-controversial-but-needed or very-urgent patches. [15:14:28] twice a day [15:15:01] halfak: Once we are in prod I'll take care of that [15:15:08] jenkins should do it [15:16:14] https://www.mediawiki.org/wiki/MediaWiki_1.27/Roadmap [15:16:26] https://www.mediawiki.org/wiki/MediaWiki_1.27/wmf.16 [15:36:26] halfak: overall what do you think of this deployment system? [15:37:09] It seems to make sense. I'd like to only adopt parts of it as we need it though. It seems heavy, but robust. [15:37:55] I would appreciate if you write a proposal (in a phab card) and we discuss about it [15:46:14] Amir1, heh. Maybe if you add a card about writing up a structured deployment process for ORES? [15:46:20] * halfak has started his meetings for the day. [15:47:51] okay [15:48:13] talking with relang people wrt several restarts [15:54:39] afk for some food, coffee [17:49:40] IT IS WORKING [17:49:46] OMG IT IS WORKING [17:50:56] https://usercontent.irccloud-cdn.com/file/r14e37iq/ [17:53:37] http://i1.kym-cdn.com/entries/icons/original/000/006/725/desk_flip.jpg [18:00:55] \o/ [18:01:00] Amir1, what's a good link to send people to our wheels repo [18:01:04] Security wants to review it [18:01:55] oh, github.com/wikimedia/research-ores-wheels [18:02:01] I typed it [18:02:04] double check [18:02:35] This is a mirror of a repo in gerrit, right? [18:02:39] Is there a gerrit view of it? [18:03:47] yeah [18:03:56] let me find the gerrit [18:04:16] (gerrit doesn't provide checking files, git.wikimedia.org and phab does) [18:05:59] Was trying to find it in git.wikimedia.org [18:06:08] * halfak conflates gerrit and git.wikimedia.org [18:07:33] https://gerrit.wikimedia.org/r/#/admin/projects/research/ores/wheels [18:08:06] Heh. OK. That's what I got to as well. It was the best I could do. [18:08:27] Why is it that our git system does not allow you to ACTUALLY LOOK AT THE FILES IN THE REPO OR AT LEAST THE README OMG [18:08:30] ? [18:08:48] halfak: http://git.wikimedia.org/summary/?r=research/ores/wheels.git [18:09:15] Differntial is much better [18:09:29] OMG! It works. [18:09:34] I hope it deploys asap [18:09:35] I'll use this. Thanks. [18:09:41] :) [18:12:43] brb lunch! [18:13:21] I need to go to home, back in ten minutes [18:56:38] that's me [19:00:53] o/ arlolra [19:01:21] I'm working on the same PR :P [19:01:49] ha, ok, was just trying to lighten your load [19:02:26] Heh no worries. [19:02:34] I was thinking that we should actually go the other way. [19:02:39] cache --> caches [19:02:54] Sorry I miscommunicated. [19:03:21] Once you pointed out that the Extractor interface called for "caches" rather than "cache", I realized that we should probably just implement that in api.Extractor [19:03:42] ah, no worries [19:03:49] happy to abandon [19:03:55] string replace took all of 2 seconds there [19:18:47] arlolra, just running some tests now to make sure I didn't break anything [19:19:06] great, but there's no rush [19:23:02] Will merge as soon as travis says OK. [19:24:11] Now pushing a new version to pypi [19:24:14] Will be 1.1.3 [19:24:28] arlolra, can you update the requirements.txt in ores to point to 1.1.3 then? [19:24:56] yup [19:27:25] wiki-ai/revscoring#590 (caches_switch - dabf00b : halfak): The build passed. https://travis-ci.org/wiki-ai/revscoring/builds/116727265 [19:27:54] OK. 1.1.3 is in pypi [19:28:28] BTW, I'm really excited that you've been looking at this arlolra. I have some fun ideas for playing with injection patterns at the Jerusalem hackathon [19:28:53] E.g. you can put on "anon goggles" and look at how edits are scored if the model thinks they were all saved by anons. [19:32:29] fun [19:32:39] sadly, I won't be in Jerusalem with you [19:32:56] pushed the version bump to my branch [19:33:06] let's see if travis likes it [19:33:20] Great. [19:33:37] I made a task for this to legitimize the work :) https://phabricator.wikimedia.org/T130250 [19:35:51] thanks [19:37:46] * halfak manually tests [19:43:26] Hmmm... Doesn't seem to actually be working. [19:43:40] Did you try sending some caches through? [19:48:29] admittedly, no. I was only testing that the data was passed in correctly. can I even test features with the revid scorer? [19:49:07] The only "feature" it uses is the rev_id and that gets overwritten in the cache. [19:49:10] I'm looking into it now. [19:49:16] I think the bug might be elsewhere. [19:50:48] right [19:51:18] I should probably setup and understand a more elaborate scorer so I can help [19:52:26] if you don't have time for this now, leave it. I'll dig in a little deeper [19:52:54] kk [19:53:00] Gotta move onto the next thing. [19:53:28] Will be back to this in a couple of hours, so leave me an update if you move into something else. [19:54:43] i'm going out. but will pick it up again tonight if you don't beat me to it. I'll keep you posted in the PR if/when I make progress [19:54:48] thakns for your time so far [21:18:32] halfak: translatewiki is going sky rocket, we have a sitenotice [21:18:48] https://translatewiki.net/w/i.php?title=MediaWiki%3ASitenotice&action=historysubmit&type=revision&diff=6741803&oldid=6729368 [21:19:38] Woah! Cool :) [21:19:53] BTW, I just added 50 new tasks to the phab board for all of our planned outreach efforts. [21:20:10] woah [21:20:16] I still have about 50 more to add :/ [21:20:31] But these tasks might get taken on by WMF CLs or our subject recruiter on the research team. [21:21:06] * halfak is doing a documentation sprint [21:22:18] cool [21:22:32] I'm waiting that ops package scap3 and release it [21:37:35] halfak: these are translations made to our project: https://translatewiki.net/w/i.php?title=Special:RecentChanges&limit=500&namespace=1206&translations=only [21:37:51] in uk, cs, fr, he, hi, id, de, and so many other [21:38:12] "Wiki-ai-wikilabels-form-dn-music-choice-label"??? [21:38:14] ps, es, it [21:39:51] halfak: https://translatewiki.net/wiki/Wikimedia:Wiki-ai-wikilabels-form-dn-music-choice-label/en [21:39:59] I think this is for edit_type [21:40:15] I don't think we have a label anywhere called "music". [21:40:29] Something broken? [21:40:34] no [21:40:38] it's for notability [21:40:55] https://www.irccloud.com/pastebin/TiFEgwid/ [21:40:55] https://www.irccloud.com/pastebin/pkQoTGQn/ [21:41:36] Amir1, Oh! I forgot. We never ran a campaign with that form. [21:41:38] Great :) [21:41:58] we haven't even ran a campaign with that form [21:42:01] :D [21:42:03] We will run a campaign soon. [21:50:41] hey halfak, I'm user vituzzu on phab [21:50:55] Hi Vito! [21:51:01] I'm currently testing labels [21:51:03] I thought we had previously solved this problem. [21:51:12] But it shouldn't be any trouble to look into it again. [21:51:24] You should see an error that says "the revision was deleted" [21:51:51] currently I'm flagging revs opening them in another tab [21:52:03] instead of skipping [21:52:46] anyway I was wondering about a different way of testing, dunno if you already took it into consideration [21:53:19] I mean, adding one or two button to any diff seen by testers would allow them to flag much more revs [21:53:31] Vito, what kind of buttons? [21:54:29] https://it.wikipedia.org/w/index.php?title=Ferdinando_Magellano&curid=71444&diff=79606595&oldid=79606567 <-- I'd like to have a button on the right which allows me to flag the rev as vandalism [21:56:05] Vito, oh! You mean in the regular MediaWiki interface? [21:56:06] more revs should avoid bias anyway [21:56:19] even a tool for testers [21:56:22] But that would cause a selection bias. [21:56:36] It's nice that, with wikilabels, we can provide random samples. [21:57:07] imho we will be able to generate a sufficient amount of data to reduce selection bias [21:58:54] though I didn't read your scientific stuffs ^^ [21:59:10] so I didn't dig much into the model are you using [21:59:36] currently I'm just Guinea pig [22:02:41] :) I think it is a good idea, but maybe hard to work out in practice. [22:02:51] I'm not against it, but it sounds like a lot of work to validate. [22:03:09] Would be happy to work with someone who wanted to lead an effort to extend the labeling system in this way. [22:07:45] I go sleep halfak [22:07:55] see you tomorrow [22:07:58] o/ Amir1|afk [22:08:12] assign some stuf to me [22:08:26] Amir1|afk, will try. You go to bed around the time I leave work :P [22:08:26] email or phab [22:08:41] * halfak hopes for an early end to his last meeting. [22:09:00] halfak: I'll spend a few more months typing weird stuffs on my matlab console, so want my contribution to your project would be the feedback and testing of an evil patroller ^^ [22:09:20] thanks [22:09:28] o/ [22:09:29] Vito: evil patroller? [22:14:52] halfak: I beat up trolls, vandals and paid editors [22:50:54] Oh. Gotcha :) [22:51:02] * halfak is in meeting mode :( [22:51:06] Only partially present