[15:00:21] hi! [15:02:31] hi! [15:02:40] I'm waiting for halfak :) [15:02:50] o/ [15:02:54] Waiting for what? [15:03:00] for the meeting [15:03:05] Oh. Want me to start it? [15:03:34] I don't know depends on awight [15:03:39] please--I haven't used this software in the longest, and don't have everyone's address [15:08:42] https://phabricator.wikimedia.org/T106838 [15:09:18] https://etherpad.wikimedia.org/p/ORES_bias [16:03:42] o/ [16:05:11] halfak: fyi: [16:05:15] https://www.irccloud.com/pastebin/bZN4sEa9/ [16:05:55] we have to merge this but it won't be a big problem https://gerrit.wikimedia.org/r/#/c/232921 [16:08:10] Cool! [16:08:19] It will be easy to merge later :) [16:08:33] Amir1, I forget what we needed to work out before I head out. [16:08:52] halfak: I want the training set [16:08:56] from PRES [16:08:59] *ORES [16:09:04] Oh yes. [16:09:17] So, have you logged into ore-compute? [16:09:20] *ores-compute [16:09:52] very long time ago [16:09:56] I forgot [16:10:16] OK. I'll try something else then. [16:11:12] I think settings is in my computer [16:11:21] so I just might need to login [16:11:28] (I'm not sure though) [16:11:50] Give it a try [16:13:05] sure [16:13:31] * halfak checks file-sizes [16:13:54] Looks like the feature sets are ~8MB uncompressed. [16:14:33] * halfak compresses them [16:14:51] OK. So you'll need to look at the feature sets in ores-wikimedia-config in order to know the column order. [16:15:19] Yeah. These are 1.5MB compressed, so they'll be easy to send along. [16:15:26] * halfak downloads. [16:16:35] nope [16:16:40] no success [16:16:58] oh thanks [16:16:59] neat [16:18:16] * halfak waits for the uploads [16:20:52] halfak: in the mean time, handling wikibase is really tricky [16:21:06] e.g. output of this and this is different [16:21:08] I can imagine. Is it mostly solvable in pywikibase? [16:21:11] https://www.wikidata.org/w/api.php?action=wbgetentities&ids=Q42&props=claims [16:21:20] https://www.wikidata.org/w/api.php?action=query&prop=revisions&titles=Q42&rvprop=content [16:21:30] not all, but enough [16:21:56] Oh. SO one of those has encoded json. [16:22:02] https://pypi.python.org/pypi/pywikibase/0.0.1 [16:22:08] But the encoded json -- does it match the regular json? [16:22:28] it is still different if you change them to dictionary [16:22:38] using json.loads() [16:23:15] 26 downloads in the last day [16:23:15] 156 downloads in the last week [16:23:15] 156 downloads in the last month [16:23:19] awesome :) [16:24:04] OK. You should have mail with the feature sets and link to the spot in ores-wikimedia-config to look for the order of features. [16:24:05] halfak: Are you sending email or uploading into somewhere? [16:24:13] oh I got the answer [16:24:22] thanks [16:24:24] :) [16:25:14] * Amir1 curses at Google Inbox [16:25:23] poor attachment handling [16:25:55] I'm going to run off. I'll be on tomorrow and doing a bit more hacking then. [16:26:08] I should have the final bits of the revscoring language refactor ready for review. [16:26:29] halfak: Ok, see you :) thanks and I will finish huge parts of features tomorrow [16:26:41] Amir1 & ToAruShiroiNeko, I'd appreciate if you could block some time tomorrow or early next week to look at it and give me notes or merges :D [16:26:50] Amir1, cool! Looking forward to it. [16:26:56] We'll pass up Stefan and Martin ;) [16:26:57] o/ [16:27:09] :) [16:27:10] o/ [21:14:33] hello