[01:12:47] 10Revision-Scoring-As-A-Service-Backlog, 10ORES, 06Operations: [spec] Active-active setup for ORES across datacenters (eqiad, codfw) - https://phabricator.wikimedia.org/T159615#3120524 (10Halfak) @mobrobac, I can't see how this would solve any of the problems we've been discussing. Can you clarify what, exa... [01:15:16] Sorry to miss you Amir1, I'd just left my desk when you asked. Tomorrow deployment sounds good. :) [01:40:35] 10Revision-Scoring-As-A-Service-Backlog, 10ORES, 13Patch-For-Review: [Spec] Tracking and blocking specific IP/user-agent combinations - https://phabricator.wikimedia.org/T137962#3120569 (10BBlack) Re: 304-vs-200, I was able to get some 304s, but only when I dropped the INM and relied on IMS. It seems like t... [02:01:15] (03PS4) 10Catrope: Add conflicts for category changes [extensions/ORES] - 10https://gerrit.wikimedia.org/r/343783 (https://phabricator.wikimedia.org/T160803) (owner: 10Mooeypoo) [02:01:50] (03CR) 10Catrope: [C: 032] Add conflicts for category changes [extensions/ORES] - 10https://gerrit.wikimedia.org/r/343783 (https://phabricator.wikimedia.org/T160803) (owner: 10Mooeypoo) [02:05:50] (03Merged) 10jenkins-bot: Add conflicts for category changes [extensions/ORES] - 10https://gerrit.wikimedia.org/r/343783 (https://phabricator.wikimedia.org/T160803) (owner: 10Mooeypoo) [11:59:23] 10Revision-Scoring-As-A-Service-Backlog, 10Wikilabels: Revisions not found should be skipped - https://phabricator.wikimedia.org/T161102#3121544 (10Strainu) [12:23:25] 10Revision-Scoring-As-A-Service-Backlog, 10Wikilabels: Hidden revisions should be skipped - https://phabricator.wikimedia.org/T161102#3121637 (10Strainu) [12:24:07] 10Revision-Scoring-As-A-Service-Backlog, 10Wikilabels: Revisions that cannot be retrieved should be skipped - https://phabricator.wikimedia.org/T161102#3121544 (10Strainu) [12:25:27] 10Revision-Scoring-As-A-Service-Backlog, 10Wikilabels: Hidden revisions should be skipped - https://phabricator.wikimedia.org/T161104#3121644 (10Strainu) [12:26:16] 10Revision-Scoring-As-A-Service-Backlog, 10Wikilabels: Revisions that cannot be retrieved should be skipped - https://phabricator.wikimedia.org/T161102#3121544 (10Strainu) [14:30:16] o/ [14:38:11] o/ [14:38:20] I need to be afk for an hour or so [15:30:54] FYI: https://phabricator.wikimedia.org/T159838#3122125 [15:31:04] I just write up an assessment of GPUs for tensor flow work. [15:31:14] Amir1, when you get back, you might be interested in ^ [15:32:10] 10Revision-Scoring-As-A-Service-Backlog, 10ORES, 06Operations: [spec] Active-active setup for ORES across datacenters (eqiad, codfw) - https://phabricator.wikimedia.org/T159615#3122138 (10mobrovac) Ok, so the focus of this ticket seems to be on how to have both Redis instances warmed up with the same content... [15:34:04] o/ Nettrom [15:34:24] o/ halfak [15:34:41] 10Revision-Scoring-As-A-Service-Backlog, 10ORES, 06Operations, 06Services (designing), 15User-mobrovac: [spec] Active-active setup for ORES across datacenters (eqiad, codfw) - https://phabricator.wikimedia.org/T159615#3122142 (10mobrovac) [15:34:51] Nettrom, do you think you'll be interested in tensor flow in the next 500 days or so? [15:35:44] halfak: probably [15:36:09] or… most likely [15:37:00] Could you take a look at the concerns raised at https://phabricator.wikimedia.org/T159838#3122125 [15:37:14] It seems that this will be an issue that us modelers should be aware of. [15:37:57] Essentially, we're deciding to either do things the way we know works but sell our open source souls or to potentially suffer massive labor-hour losses in order to preserve our open source souls. [15:38:12] Works but closed source drivers == Nvidia [15:38:23] Experiment with open source drivers == AMD [15:38:55] oh, interesting… reading the task conversation now [15:53:57] I remember Linus Torvalds saying [15:54:08] '[CENSORED] you, nvidia!' [15:55:33] anyway that's interesting discussion [16:03:08] revi, :) sounds like a Linus quote [16:03:19] it is :D [16:05:03] halfak: interesting readings about the development… from https://github.com/tensorflow/tensorflow/issues/22 (linked from the SO answer you linked to), it sounds like there won’t be any official support any time soon, meaning it’s all experimental stuff [16:05:29] Nettrom, right. Seems like the docs on the experimental support suggest that things actually do "work" [16:10:02] halfak: for some definition of “work”, I’m skeptical of anything that can’t be installed with apt-get or pip, wondering how much effort it will require by ops to get it to work [16:10:38] I’m wondering if it’s mainly a question of how soon we need TF, if it’s _now_ then Nvidia is the way to go, if it’s “maybe soon” then maybe not? [16:11:52] "Amir, the developer of ORES" https://lists.wikimedia.org/pipermail/wiki-research-l/2017-March/005734.html [16:12:07] Amir1, do you want to correct GerardM or should I? [16:28:02] * Nettrom relocates, AKF [16:45:02] halfak: Is that the email from GerardM that you told me today? [16:45:26] Na. I've been getting personal emails from him. If he continues, I might report him. [16:46:22] halfak: what? he has written you emails for a while already? [16:46:37] that is hilarious [16:47:06] yeah. He's one of those people who hits you through all mediums at once. I'm surprised he's not trolling us here too. [16:47:47] halfak: I think it is because he does not that we are in this channel :) [16:48:16] does not know* [16:48:28] no one tell him [16:48:29] :D [16:50:11] halfak: Hey, I was afk, what's up [16:50:27] GerardM :) [16:50:33] Amir1: you were mentioned in email to halfak [17:29:37] * awight emits concerned looks from remote peanut gallery [17:39:46] lol "actionable difference of at least 2%" is fantastic [17:56:13] awight, I'm glad you find that absurd too. :) [20:23:10] glorian_wd, you asked what the purpose of labeling an item twice is. [20:23:20] It so we can find out if people consistently label items the same. :_) [20:23:20] yes [20:23:41] what will happen if an item is labeled as "A" and "D"? [20:24:01] perhaps, we wont use such item for the model? [20:24:09] Then we'll make a post on the talk page about it and ping the people who labeled it. [20:24:33] We might need to clarify or update the criteria to make it easier to be consistent. [20:24:34] did that happen often in the past? [20:25:32] Past? [20:25:42] in the past ORES campaign [20:25:54] such inconsistencies [20:44:57] halfak: ? [20:47:00] glorian_wd, yeah. With edit_type [20:47:04] We did 3 pilots. [20:47:17] https://en.wikipedia.org/wiki/Wikipedia:Labels/Edit_types [22:13:32] 10Revision-Scoring-As-A-Service-Backlog, 10Education-Program-Dashboard, 10ORES, 10Outreach-Programs-Projects, and 4 others: Add automatic article feedback feature to Wiki Ed Dashboard / Programs & Events Dashboard - https://phabricator.wikimedia.org/T158679#3123473 (10Halfak) [22:31:32] 06Revision-Scoring-As-A-Service, 10rsaas-articlequality : [Discuss] Hosting the monthly article quality dataset on labsDB - https://phabricator.wikimedia.org/T146718#3123544 (10yuvipanda) a:05yuvipanda>03None [22:32:55] 06Revision-Scoring-As-A-Service, 10rsaas-articlequality : [Discuss] Hosting the monthly article quality dataset on labsDB - https://phabricator.wikimedia.org/T146718#3123555 (10Halfak) @yuvipanda, thanks for helping with the workshop. o/ [22:38:00] halfak: Everything is deployed now, RTL and LTR looks good \o/ [22:39:19] \o/ [22:39:22] Nice work!