[06:50:12] wikimedia/articlequality#250 (euwiki_match_eu-en-es - 213b3c8 : KEVIN BAZIRA): The build passed. https://travis-ci.org/wikimedia/articlequality/builds/607501061 [11:18:20] fyi, the gap in https://grafana.wikimedia.org/d/HIRrxQ6mk/ores?orgId=1&from=1572941180448&to=1572951519648 is because of an effort to move forward the statsd export to the prometheus exporter. I 've had to revert it due to it [14:30:17] o/ kevinbazira [14:30:30] How's model tuning going? [14:31:59] Checking emails now [14:33:19] halfak o/ [14:33:48] The model tuning is going well. Sent details via email. [14:34:01] Can you add those tuning reports to the PR? [14:39:50] FWIW, those kinds of warnings are normal. Sometimes various model/param combinations are just too slow to train, so we cut them off at an hour. [14:42:53] Looks like we can get some better performance out of the euwiki model by changing some of the hyperparams (model building params/learning params). [14:43:23] A RandomForest wins, but not by too much and it tends to take up extra memory. Still, I want to build one and see how big the model file is. [14:44:24] kevinbazira, you can take the values from the tuning report and copy them into the Makefile. [14:44:33] RandomForestClassifier | min_samples_leaf=3, max_features="log2", n_estimators=320, criterion="gini" [14:45:19] Alright [14:52:23] There's something funky with your configuration for git-lfs. [14:52:35] Are you submitting your changes directly from stat1007? [14:54:39] Yep [14:59:51] Hmm. It should just work. I just checked your local repo and it has the right config. [15:00:05] Which it should since the config comes with the repo. [15:00:16] I'll need to dig into this more later. I'm running to a meeting. [15:55:10] back! [16:58:10] emails sent. [16:58:18] Async time! [17:03:47] Y: I worked on session-orientation. I have solved the listy-functiony problem. I did it using a decorator that denotes a method as a "meta-dependency" and then applying another decorator to all meta-dependencies in the list_of_tree method. I've almost squashed the last bug. I expect to get tests passing today and start work on the extractor. I'll have a report about how things work for Andy to review. Otherwise, I drafted a set of [17:03:47] instructions for Kevin to make progress. I also reached out to OIT again about shipping Kevin's laptop. Seems like they are not keen on paying the $400 import tax. [17:03:47] T: Lots more meetings today. I did some work with Kevin to help him get started with the next steps. Looks like our euwiki model has potential. I haven't seen the results of the tuning report for cawiki though. I'll continue work on session-orientation if I get time. I'm also starting some meetings with the Growth team to talk to them about using our topic models. These are more planning meetings as they have already decided to adopt ORES [17:03:50] for their products in this space (wooo!). [17:11:39] Y: I fixed a couple more things on the Jade API patchset: percent-encoding example urls, and also DRY'd up the API modules with a new base class. Responded to the latest review in gerrit and also setup time for a walthrough of the API on Thursday. [17:12:13] T: Planning to continue working on the Jade UI "Propose New Label" workflow. The goal is go get a OOJSUI process dialog box rendering a form that we can use to send data to our CreateAndEndorse API module. [17:13:40] Y: Moved changes I'd made on the Catalan model to a branch on the main repo as advised by Aaron. PR: https://github.com/wikimedia/editquality/pull/216 [17:13:53] Y: Reviewed Aaron's PR on that adds new data for the euwiki model https://github.com/wikimedia/articlequality/pull/83 [17:14:08] T: Tuned and Rebuilt Basque models with words that match eu, en, es dicts. PR: https://github.com/wikimedia/articlequality/pull/98 [17:14:21] T: Tuned Catalan models and waiting to review reports with Aaron before I make PR. [17:22:21] Re that catalan tuning, I think we need the tuning reports in the PR before we can make any decisions about it. [17:22:30] kevinbazira, can you run "git lfs --help" on stat1007? [17:22:46] I'm curious if you environment has access to the lfs module of git. [17:23:06] etherpad seems to be down :( [17:23:10] Alirght, I'm let me run it [17:23:31] Etherpad is back! [17:25:04] "git lfs --help" run fine on stat1007 [17:41:09] I don't know why the LFS files aren't getting picked up. [18:04:25] OK. Tests are passing on session-oriented stuff. I just pushed my changes. Once travis is happy, I'll write a report. [18:04:30] But for now, I'm off to lunch! [18:04:39] And... I just saw an icinga ping. [18:04:43] So I guess I'm still here. [18:05:18] nevermind. It's just wmflabs and the warning is for the ssh cert. [18:05:46] Looks like -cloud knows about it. [18:05:55] https://phabricator.wikimedia.org/T237066 [18:06:00] OK really going to lunch [20:48:09] lunchin