[07:59:56] 10Scoring-platform-team (Current), 10editquality-modeling, 10User-Ladsgroup, 10User-Tgr, 10artificial-intelligence: Train/test damaging and goodfaith model for Hungarian Wikipedia - https://phabricator.wikimedia.org/T185903#4132791 (10awight) @Bencemac Thanks for the nudge, I need to update this task to... [11:20:58] 10Scoring-platform-team, 10articlequality-modeling, 10draftquality-modeling, 10drafttopic-modeling, 10artificial-intelligence: Enable ores extension article quality storage in beta cluster - https://phabricator.wikimedia.org/T192267#4133192 (10Ladsgroup) [11:21:05] 10Scoring-platform-team, 10articlequality-modeling, 10draftquality-modeling, 10drafttopic-modeling, 10artificial-intelligence: Enable ores extension wp10 storage in English Wikipedia - https://phabricator.wikimedia.org/T192268#4133202 (10Ladsgroup) [12:56:22] 10Scoring-platform-team, 10editquality-modeling, 10artificial-intelligence: [research] Why is the japanese 'reverted' model so bad? - https://phabricator.wikimedia.org/T133405#4133407 (10Ladsgroup) If you think so, I'm fine :D [14:16:15] o/ [14:20:11] hi halfak [14:20:47] Hi vermont! [14:21:09] Do you happen to know if Wikipedia holds cryptocurrencies? [14:22:08] Uh. What? [14:23:11] 10Scoring-platform-team: Drafttopic: Add utility to extract dependents - https://phabricator.wikimedia.org/T183392#3851910 (10Halfak) Is this still relevant? [14:23:38] 10Scoring-platform-team: Drafttopic: add article text fetching utility - https://phabricator.wikimedia.org/T183355#3851037 (10Halfak) Is this still relavant? [14:23:41] https://wikimediafoundation.org/wiki/Ways_to_Give#bitcoin if that helps [14:24:14] 10Scoring-platform-team (Current), 10Research Ideas, 10drafttopic-modeling, 10artificial-intelligence: Revscoring tune does not recognize a set of labels as target - https://phabricator.wikimedia.org/T181163#4133712 (10Halfak) Is [14:24:26] 10Scoring-platform-team (Current), 10Research Ideas, 10drafttopic-modeling, 10artificial-intelligence: Revscoring: Statistic for multilabel classification - https://phabricator.wikimedia.org/T181166#4133717 (10Halfak) [14:24:29] 10Scoring-platform-team (Current), 10Research Ideas, 10drafttopic-modeling, 10artificial-intelligence: Revscoring tune does not recognize a set of labels as target - https://phabricator.wikimedia.org/T181163#4133714 (10Halfak) 05Open>03Resolved a:03Halfak [14:24:59] Vermont, see what apergos said. is that what you are looking for? [14:25:39] yep thanks [14:26:36] 10Scoring-platform-team (Current), 10Research Ideas, 10artificial-intelligence: Train/test draft topic model (new article routing AI) - https://phabricator.wikimedia.org/T123327#4133746 (10Halfak) [14:26:52] 10Scoring-platform-team (Current), 10Research Ideas, 10artificial-intelligence: Train/test draft topic model (new article routing AI) - https://phabricator.wikimedia.org/T123327#1926288 (10Halfak) This was done a while ago. We're working on deployment now. [14:27:02] 10Scoring-platform-team (Current), 10ORES, 10artificial-intelligence: Deploy topic model to ORES - https://phabricator.wikimedia.org/T176336#4133751 (10Halfak) [14:27:05] 10Scoring-platform-team (Current), 10Research Ideas, 10artificial-intelligence: Train/test draft topic model (new article routing AI) - https://phabricator.wikimedia.org/T123327#4133749 (10Halfak) 05Open>03Resolved a:03Halfak [14:27:25] 10Scoring-platform-team (Current), 10ORES, 10drafttopic-modeling, 10artificial-intelligence: Deploy drafttopic model to ORES - https://phabricator.wikimedia.org/T176336#3621858 (10Halfak) [14:30:51] 10Scoring-platform-team (Current), 10ORES, 10drafttopic-modeling, 10artificial-intelligence: Check drafttopic model memory usage - https://phabricator.wikimedia.org/T192293#4133770 (10Halfak) p:05Triage>03Normal [15:02:40] Amir1, ewhit: Meeting time! [15:02:48] * halfak looks for awight [15:03:08] I think I rsvp'ed no to this? Sorry, I'm in a class right now! [15:05:22] Gotcha. [15:06:18] brb getting snacks while I wait for people to show up. [15:09:46] halfak: heads-up that I’m working today, but only after a terrible errand this morning. [15:10:01] So no sync meeting? [15:10:03] U in the sync meeting / want to chat? [15:10:13] ok headed, my mistake [15:59:13] skipping bs bec of dinner :P [16:01:54] kk [16:32:31] I need to do this terrible errand statim, o/ [16:40:06] halfak: the abstract section is only 150 words :/ [16:40:12] *restricted to [17:06:56] codezee, always do-able. [17:24:25] * halfak --> Lunch [17:39:38] halfak: does the abstract in the portal and one in the paper need to be same?( 150 words?) [18:23:58] I’m going to deploy new huwiki models to beta, for fun... [18:34:39] surprise lunch meeting [18:34:40] :( [18:34:58] codezee, not sure. But 150 words is a good limit. We should be able to trim [18:51:44] (03PS1) 10Awight: Update editquality submodule [services/ores/deploy] - 10https://gerrit.wikimedia.org/r/426967 [18:52:20] (03CR) 10Awight: [V: 032 C: 032] "merging own submodule bump." [services/ores/deploy] - 10https://gerrit.wikimedia.org/r/426967 (owner: 10Awight) [19:07:14] awight, I'm just about to restart the process of re-training everything for revscoring 2.2.0 [19:07:26] did the wikimedia bastion change? [19:07:33] I'm timing out against bast1001 [19:09:12] looks like it's bast1002 now maybe? [19:09:52] confirmed [19:11:24] Looks like I should cut 2.2.1 first [19:12:38] {{done}} [19:12:39] How cool, halfak! [19:13:30] codezee, what are our expected scipy/numpy versions? [19:13:34] for gensim? [19:13:47] I'm getting scipy 0.17.1 and numpy 1.10.4 [19:13:50] Is that right? [19:14:15] gensim of 3.3.0 [19:15:01] 10Scoring-platform-team, 10ORES: ORES beta will need more disk space - https://phabricator.wikimedia.org/T192311#4134487 (10awight) [19:16:27] oops, reading backscroll [19:16:33] yes, it was a hard cutover to bast1002 [19:17:52] halfak: afaik scipy should be > 1.0.0 [19:17:58] checking [19:18:18] we don't have that in requirements. I assume gensim *should* have had it in its requirements? [19:19:08] https://github.com/RaRe-Technologies/gensim/blob/develop/setup.py#L309 [19:19:21] Looks like the current version requires >= 0.18.1 [19:20:50] halfak: i remember now, thats the min requirements, but we decided to do scipy and numpy upgradate to latest as that was long overdue - https://github.com/wiki-ai/revscoring/pull/391/files [19:21:05] *upgrade [19:21:10] OK we need to put that into revscoring's requirements.txt [19:21:23] (03PS1) 10Awight: Reduce cached deployment revisions to 1 [services/ores/deploy] - 10https://gerrit.wikimedia.org/r/426972 [19:21:28] We should bump the min requirements -- not just the max :) [19:21:54] halfak: oh, ok i get, it must be happening that its installing a lower version [19:22:06] Or just keeping the lower version [19:22:18] I'll get a PR quick [19:25:07] Since we're getting out of the range of compatibility, we might as well upgade scikitlearn too. [19:28:00] wiki-ai/revscoring#1468 (req_restrict - 34bf015 : halfak): The build passed. https://travis-ci.org/wiki-ai/revscoring/builds/367320548 [19:29:17] I now have numpy-1.14.2 scikit-learn-0.19.1 scipy-1.0.1 [19:30:13] codezee or awight: got a quick review for me? https://github.com/wiki-ai/revscoring/pull/397 [19:30:52] wiki-ai/revscoring#1470 (req_restrict - b1ce22c : halfak): The build passed. https://travis-ci.org/wiki-ai/revscoring/builds/367321500 [19:32:57] {{done}} [19:32:57] How cool, awight! [19:33:11] Thnks [19:34:14] * halfak releases revscoring-2.2.2 [19:34:54] that was quick ... :D [19:40:49] 10Scoring-platform-team, 10Performance: Investigate ORES performant window in mid-August, 2017 - https://phabricator.wikimedia.org/T175640#4134520 (10awight) 05Open>03declined [19:43:30] I can build the wheel [19:43:51] Rock on awight. [19:43:55] I'll be focusing on the models. [19:44:11] I thought you were writing a paper! [19:44:26] Figured I could kick this off. [19:44:31] O_O [19:44:33] I want to get a clean ORES deploy out soon :) [19:44:46] I could leave the model rebuild to you too :P [19:44:54] ah perfect, actually [19:45:15] I’m blocked by the version incompatibilities anyways. [19:48:14] OK I just hit "go" on model rebuilding. I'll get back to writing now :) [19:52:21] wiki-ai/editquality#301 (revscoring-2.2.2 - e6c68d0 : Adam Wight): The build passed. https://travis-ci.org/wiki-ai/editquality/builds/367331754 [19:53:40] halfak: You’re rebuilding editquality, right? I can build the others, then. [19:53:55] Yup. I'll focus on editquality. [19:59:05] wiki-ai/articlequality#60 (revscoring-2.2.2 - e831292 : Adam Wight): The build passed. https://travis-ci.org/wiki-ai/articlequality/builds/367334697 [20:04:55] (03CR) 10Halfak: [V: 032 C: 032] Reduce cached deployment revisions to 1 [services/ores/deploy] - 10https://gerrit.wikimedia.org/r/426972 (owner: 10Awight) [20:12:01] 10Scoring-platform-team (Current), 10ORES: Make it easier to keep production and labs deployments in sync - https://phabricator.wikimedia.org/T181824#4134616 (10awight) This is really annoying me today. Moving to the active board... [20:18:41] halfak: the abstract overshot a few words beyond 150 but the portal didn't complain while saving. That should be ok? or we should be precise? [20:19:38] Hmm. Probably doesn't matter, but a few words would be easy to deal with. [20:19:58] Ha. Hit 159 words with the ORES paper abstract [20:20:04] haha :P [20:21:20] I like portals like easychair, which are clearly doing no validation whatsoever [20:21:38] built in php I’m sure, and still chugging along [20:22:35] makes me hope that the the pendulum will eventually swing back towards “liberal with inputs” in UX [20:22:39] like the good old Wikipedia ;) [20:22:48] Intelligent algorithms have a long history of making curation work in peer production tractable. From counter-vandalism to task routing, basic machine prediction allow open knowledge projects like Wikipedia scale to the largest encyclopedia in the world. However, the ideologies and values of the community were captured in the development of these algorithms and the processes they support. Wikipedia's challenges and the community's values [20:22:48] have changed in the last decade, but its algorithmic support systems have remained largely stagnant. The conversation about what quality control should be and what place algorithms have remains restricted to a few expert engineers. In this paper, we describe ORES: an algorithmic service designed to open up socio-technical conversations in Wikipedia to a broader set of participants. In this paper, we argue the theoretical mechanisms of social [20:22:49] change ORES enables and we describe the phenomena around ORES from the 3 years since ORES' deployment. [20:23:45] codezee: hehe, yeah meanwhile I just wrote one of the most thorough input validators of my career just to be a jerk to 3rd-party tool authors. [20:26:00] halfak: finally got it below the belt or leaving at 159? [20:27:34] Below the belt :) [20:33:10] I'm dragging my feet. Gonna go lay down for a bit and try to get back to writing. [20:46:55] (03PS1) 10Awight: Update wheels for revscoring 2.2.2 [research/ores/wheels] - 10https://gerrit.wikimedia.org/r/427015 [20:53:31] (03PS1) 10Awight: Rebuild wheels for revscoring 2.2.2 [services/ores/deploy] - 10https://gerrit.wikimedia.org/r/427017 [21:01:13] halAFK: brought it down to 148 somehow, at the same time preserving the points. You can take a look on overleaf if its sane and maybe also take a glance at the intro sections [21:01:34] the latter one's are mostly statistics, I'll take another pass tomorrow to remove any errors [21:01:52] for today I've addressed the points that were discussed in the sync [21:19:08] wiki-ai/ores#935 (revscoring-2.2.2-2 - 821923d : Adam Wight): The build passed. https://travis-ci.org/wiki-ai/ores/builds/367370859 [21:22:04] wiki-ai/ores#936 (revscoring-2.2.2-2 - 7820fcb : Adam Wight): The build passed. https://travis-ci.org/wiki-ai/ores/builds/367372158 [22:20:24] Ugh. Too much writing makes me exhausted. Need to keep a pace to this. [22:28:24] codezee, thanks for getting that taken care of. [23:40:18] https://www.mediawiki.org/wiki/Topic:Ubeq2ro6ddkinvv9 [23:40:21] GOOD ENOUGH [23:40:22] lol [23:40:29] I'm out of here for the day. [23:40:32] Take care folks! [23:40:48] BTW, Adam the above link is very rough, but I think you'll like it. [23:41:01] I think it was this set of ideas that original drew you to work on this team ^_^ [23:41:18] (If I'm remembering Wikimania 2015 right) [23:44:04] :D [23:45:02] ok awesome—when I’m done sifting for uranium with a piece of chickenwire, I’ll work on responses [23:45:28] currently having a rough time with the revscoring update, it’s been a few months, there’s an exciting package rename, etc. [23:55:39] 10Scoring-platform-team (Current), 10ORES, 10articlequality-modeling, 10User-Ladsgroup, 10artificial-intelligence: Rename "wikiclass" to "articlequality" - https://phabricator.wikimedia.org/T181820#3803377 (10awight) Some more pieces we need, I'm currently fixing a few that I need for a revscoring upgrad... [23:57:48] 10Scoring-platform-team (Current), 10ORES, 10articlequality-modeling, 10User-Ladsgroup, 10artificial-intelligence: Rename "wikiclass" to "articlequality" - https://phabricator.wikimedia.org/T181820#4134900 (10awight) * articlequality/wikiclass.py needs to be renamed