[10:57:07] 10Scoring-platform-team (Current), 10articlequality-modeling, 10artificial-intelligence: Update documentation for ArticleQuality.js - https://phabricator.wikimedia.org/T209387 (10He7d3r) For future reference: there is now a translation at https://pt.wikipedia.org/wiki/User:EpochFail/ArticleQuality [14:59:10] does everyone see the new etherpad or is that just me [14:59:15] xD [14:59:42] I see it too. Changed the font to a monospace font right away [15:03:37] haha [16:10:32] 10ORES, 10Scoring-platform-team (Current): Estimate ORES CapEx for FY21 - https://phabricator.wikimedia.org/T249917 (10Halfak) I'm not quite sure how much headspace we would get by moving the uwsgi workers off servers that celery is running. I think it'll gain us a much smaller amount than doubling the RAM.... [16:27:33] 10Scoring-platform-team (Current), 10Wikilabels, 10editquality-modeling, 10artificial-intelligence: Build draft quality model for ptwikipedia - https://phabricator.wikimedia.org/T246667 (10Halfak) I was thinking... Would it be better to print the ORES prediction on the page somewhere so that it is easier... [16:30:54] 10ORES, 10Scoring-platform-team (Current): Estimate ORES CapEx for FY21 - https://phabricator.wikimedia.org/T249917 (10Halfak) Oh! Of course, we could also do some routing with celery. E.g. we can probably load up a limited set of models on some nodes. I'm not sure how complicated that would be. It's defin... [19:14:16] I'm working on a new revscoring version. This version should support python 3.5-3.8. It'll have idioms. It'll also have improved tokenization performance. [19:18:11] weee! [19:23:16] nice [19:48:48] * halfak waits patiently or travis [19:48:58] Actually that's a lie. I'm very impatient. [19:52:28] I see [19:53:55] * halfak starts working with pyenv to get 3.8 installed for testing. [19:55:22] 10ORES, 10Scoring-platform-team (Current), 10Documentation: Automate Sphinx docs for ORES repo - https://phabricator.wikimedia.org/T252173 (10ACraze) [19:55:24] 10ORES, 10Scoring-platform-team, 10Documentation, 10Epic: [Epic] Clean up ORES documentation - https://phabricator.wikimedia.org/T148974 (10ACraze) [20:03:00] * halfak waits forever for the scipy wheel to build for 3.8.3. [20:03:09] You'd think they would have a pre-built wheel for this. [20:19:21] 10Scoring-platform-team (Current), 10drafttopic-modeling: Why does loading the drafttopic models take so much memory? - https://phabricator.wikimedia.org/T250435 (10Halfak) 05Open→03Resolved I think that this task is mostly done. It seems that the topic models are as large as expected. [20:19:23] 10ORES, 10Scoring-platform-team (Current): Estimate ORES CapEx for FY21 - https://phabricator.wikimedia.org/T249917 (10Halfak) [20:41:01] 10ORES, 10Scoring-platform-team (Current), 10artificial-intelligence: Review model performance for ptwiki 'articlequality' and 'draftquality' - https://phabricator.wikimedia.org/T250809 (10He7d3r) @Halfak: Oops... I missed the `-v` flag when I used grep to remove the bot assessments. So, instead of consideri... [20:41:24] halfak, I messed up a little ^ =S [20:44:41] That's awesome. [20:45:13] I mean, it's not awesome you messed up but it's awesome that the data is sane. [20:45:29] Also that we could potentially get a performance boost from building a model on this cleaner data. [20:45:32] What do you think? [20:45:41] Also, we could add a username regex in the extractor. [20:46:01] Or just filter in the makefile using grep [20:47:10] halfak, you mean like this? https://github.com/wikimedia/articlequality/compare/master...he7d3r:since_2014 [20:48:12] Right. [20:48:48] It is essentially what I did in the command line, but we can add something like that to the Makefile [20:49:25] Right. Extra credit for including a comment block explaining why. [20:50:15] Sure. But I think I should wait for the patch travis is working on [20:50:39] and then build the model again to check its real performance [20:50:52] (on human data at this time) [20:50:58] Yeah. Travis seems to be jammed :\ [20:55:57] * halfak resorts to testing other versions of python himself. [20:57:11] 10ORES, 10Scoring-platform-team (Current), 10artificial-intelligence: Review model performance for ptwiki 'articlequality' and 'draftquality' - https://phabricator.wikimedia.org/T250809 (10Halfak) [20:57:13] 10ORES, 10Scoring-platform-team (Current), 10artificial-intelligence: Write report about misclassification reports - https://phabricator.wikimedia.org/T251905 (10Halfak) 05Open→03Resolved [21:30:04] He7d3r, finally merged! [21:30:27] In unrelated news, I updated flake8 and was rewarded with a whole pile of new warnings and errors. [21:30:29] FML [21:30:48] Wohoo! [21:38:08] And it looks like I broke documentation with a reference to "sklearn.ensemble._gb.GradientBoostingClassifie" [21:47:19] wikimedia/articlequality#385 (remove_bots - ac1c8df : Helder): The build failed. https://travis-ci.org/wikimedia/articlequality/builds/689011391 [22:09:55] halfak, correct me if I'm wrong: in order for the tests to pass after https://github.com/wikimedia/articlequality/pull/130 I would need to create dummy user classes and add the user to all existing revisions in the tests?