[00:35:48] halfak: https://medium.com/imgur-engineering/lessons-learned-writing-highly-available-code-7eaf3d7aae00 nice reading for building systems with useful tidbits for ORES [13:24:33] o/ yuvipanda, yeah was reading that. [14:05:00] ROC-AUC for fawiki "goodfaith" is 0.94! [14:21:11] o/ ToAruShiroiNeko [14:22:24] yuvipanda, what do you think of http://pythonhosted.org/mwdb/? [14:22:33] I finally figured out what I wanted it to do. [14:22:35] hello [14:22:43] ... cleanly map between the production DBs and labs. [14:23:41] ToAruShiroiNeko, I'm hacking on editquality today. I'll be moving the revert labeler there and getting the workflow for generating models together. [14:24:01] I'd like ORES to call out to this library to get feature lists, build models, etc. [14:24:09] ok [14:24:13] how can I help? [14:24:23] Nothing right now. Just letting you know what I'm hacking on. [14:24:24] OH! [14:24:31] I have lots of PRs in from yesterday. [14:24:44] See https://github.com/wiki-ai/revscoring/pull/195 [14:24:50] sure [14:24:53] that I can look into [14:25:01] I do PR and I do PR :p [14:25:11] And when you are done with that, this one depends on it. https://github.com/wiki-ai/ores/pull/97 [14:25:27] That'll get the ORES extension unblocked on their cache invalidation work. [14:29:34] WOot! [14:30:14] With that I can, (1) release a new version and (2) start building models that know about their accuracy stats. [14:30:27] The ORES one should be easy after that. [14:30:34] both merged [14:30:44] Cool. [14:30:45] review was fairly easy given the nature of change [14:30:47] Thanks, dude. [14:30:57] thanks for filing the pr :p [14:31:06] Yeah. Not much is special and the nature of it afforded a good test for the PR description [14:31:11] \o/ [14:31:36] So, I'm going to try to get the editquality library in shape for building models, but it's good to go for prelableing. [14:34:09] arite [15:08:46] https://github.com/wiki-ai/revscoring/releases/tag/0.6.4 [15:15:28] It all works! [15:15:30] Sort of. [15:15:35] Not on staging yet, but getting close. [15:17:02] We're going to have 6 new models up by EOD if I have my way. [15:17:06] Exciting stuff. [16:01:31] Arg. I really want to merge this guy before I deploy anything today. [16:01:31] https://github.com/wiki-ai/ores/pull/96 [16:01:36] yuvipanda, ^ ? [16:01:45] Any chance you can take a look at how I am using statsd [16:03:25] I'm also curious how you feel about python wheels. [16:03:43] They seem like they'd be better in a prod env since they don't require execution of code to install. [16:04:25] numpy, scipy and sklearn all support wheels. [16:07:09] * halfak experiments with compiling a bdist_wheel for sklearn [16:08:05] * halfak considers self-merging the statsd stuff. [16:08:18] It's not going to cause problems if I and just doing tests in staging. [16:08:26] Then I can figure out what our metrics are going to look like. [16:16:36] See notes here about python wheels: https://phabricator.wikimedia.org/T108556 [17:10:15] halfak should I look at https://github.com/wiki-ai/ores/pull/69 ? [17:10:46] Looks like that one might be abandoned. [17:10:49] Maybe you could take it over. [17:12:28] you mean yuvi right? [17:12:54] there is also https://github.com/wiki-ai/revscoring/pull/146 [17:12:58] Na. That's install docs. [17:13:24] Yeah. Seems like this one might have died too. https://github.com/wiki-ai/revscoring/pull/146 [17:13:39] I'd like to pick that one up at some point [17:13:44] I've been working with tox. [17:13:59] yeah [17:14:03] it has conflicts too [17:49:31] ToAruShiroiNeko, can you take a look at this: https://github.com/wiki-ai/revscoring/pull/196 [17:59:26] I added some notes to https://github.com/wiki-ai/revscoring/pull/196 [17:59:33] It shows that the ranges work as expected. [19:10:48] halfak sure [19:11:00] Too late [19:11:18] sorry I am kind of distracted with my uncles death [19:11:33] I dont know what to think [19:21:33] Hey dude. I'm sorry your struggling. You don't need to be working today. [19:21:36] *re [19:22:38] I'll just self merge for a while. :) [19:23:19] a self-merge is basically http://cdn.viralscape.com/wp-content/uploads/2014/09/Dog-Walks-Itself-On-Leash.jpg [19:23:49] Apt metaphor! [19:52:57] https://meta.wikimedia.org/wiki/Research_talk:Revision_scoring_as_a_service#Better_usage_statistics_in_graphite [19:53:21] I'm ready to deploy now, but I want to have yuvipanda nearby when I do since we changed to the submodules system. [19:55:55] thanks [19:56:03] it means a lot to me [19:56:14] past three hours have been a blur