[08:11:20] Hello! [08:15:31] Mroning! [08:15:40] (evidently, I am not awake yet...) [08:25:19] haha , good morning Tobias [08:45:37] 06Machine-Learning-Team, 13Patch-For-Review: Deploy logo-detection model-server to LiftWing staging - https://phabricator.wikimedia.org/T362749#9743875 (10mfossati) >>! In T362749#9729294, @kevinbazira wrote: > @mfossati, when a model-server is deployed within the WMF k8s infrastructure it has to be configured... [09:21:34] 06Machine-Learning-Team: Configure the logo-detection model-server hosted on LiftWing to process images from Wikimedia Commons - https://phabricator.wikimedia.org/T363449 (10kevinbazira) 03NEW [09:26:34] 06Machine-Learning-Team, 13Patch-For-Review: Deploy logo-detection model-server to LiftWing staging - https://phabricator.wikimedia.org/T362749#9744018 (10kevinbazira) Thank you for sharing an example of the public stash URL, @mfossati! In T363449, we are going to configure the logo-detection model-server host... [09:34:32] 06Machine-Learning-Team, 06serviceops, 13Patch-For-Review: Rename the envoy's uses_ingress option to sets_sni - https://phabricator.wikimedia.org/T346638#9744042 (10JMeybohm) [10:25:01] * klausman lunch [10:46:03] kevinbazira: I'm stumbling in an issue with logo detection makefile. since tensorflow-cpu isn't available for apple silicon I tried to run everything in a python3-bookworm image and I get the following error (Illegal instruction) more here -> https://phabricator.wikimedia.org/P61209 [10:46:32] I'll try to debug it later again, was just wondering if you've faced this issue [10:48:17] isaranto: o/ [10:49:03] thank you for testing the logo-detection Makefile recipe [10:53:03] sorry about the apple silicon limitation. I have not experienced this issue. I have been running it successfully using this image: docker-registry.wikimedia.org/bookworm:20240414 [10:54:11] it can be found here: https://docker-registry.wikimedia.org/bookworm/tags/ [11:02:53] yeah it is the same image jus this has python pre installed [11:03:10] I'll try again later and let you know [11:03:18] * isaranto lunch [11:15:45] I think the packaged Python is likely x86_64 only, and the M1 can't run its code SIGILL (Illegal Instruction) is the typical error/abort reason when that happens [11:18:22] klausman: o/ chiming in [11:18:22] isaranto: okok whenever you're ready to try it again I've added some notes here: https://phabricator.wikimedia.org/P61209#246335 [11:19:11] klasuman: thank you for chiming in. here is the patch in question: https://gerrit.wikimedia.org/r/c/machinelearning/liftwing/inference-services/+/1023532 [11:26:58] I'll give the "make logo-detection" approach a whirl [11:28:01] Works fine here [11:29:03] The approach from your comment on Ilias' paste also works fine. Latency is about 900ms [11:34:26] great. thank you for testing it klausman. adding you as a reviewer :) [12:07:16] (03CR) 10Klausman: [C:03+1] "LGTM, but we should wait for Ilias or Aiko to confirm this works with non-x86 machines as well." [machinelearning/liftwing/inference-services] - 10https://gerrit.wikimedia.org/r/1023532 (https://phabricator.wikimedia.org/T363294) (owner: 10Kevin Bazira) [12:08:59] isaranto: I had the same issue where tensorflow-cpu isn't available for apple silicon. I tried to install tensorflow instead and it worked [12:09:33] maybe tensorflow-cpu is missing sth necessary for running on mac os. I suggest we could add a comment in the makefile for mac users [12:10:48] kevinbazira: --^ ah I mean in readme [12:11:24] aiko: thanks! apple silicon works with standard tensorflow so no need for tensorflow-cpu . this is a different issue though. I have ran the logo detection in the past locally in previous reviews but this time I was trying to run it in a clean docker image env so that I could do it from scratch [12:12:21] aiko: o/ hope you're feeling better today. thank you for testing. I am going to add a note to the readme :) [12:13:08] sry my bad it is not a different issue. But I figured that inside the container it wont have an issue [12:16:03] thanks for the help everyone I'll report back on this but we don't need to worry about this, the patch is fine! [12:18:08] How do people feel about reorganizing that Makefile a little (after Kevin has merged his change)? Som targets that should be marked phony aren't, and I'd order the targets for the various models alphabetically and have run/run-server and clean at the bottom [12:20:20] sounds good! [12:21:29] kevinbazira: thanks! [12:23:04] klausman: sounds good +1 [12:23:28] (03PS2) 10Kevin Bazira: Makefile: add support for logo-detection [machinelearning/liftwing/inference-services] - 10https://gerrit.wikimedia.org/r/1023532 (https://phabricator.wikimedia.org/T363294) [12:24:43] --^ pushed a patch set with a note for mac users in the readme file [12:24:54] (03CR) 10AikoChou: [C:03+1] Makefile: add support for logo-detection [machinelearning/liftwing/inference-services] - 10https://gerrit.wikimedia.org/r/1023532 (https://phabricator.wikimedia.org/T363294) (owner: 10Kevin Bazira) [12:26:26] klausman: +1 on reorganizing the Makefile [12:36:16] (03CR) 10Kevin Bazira: [C:03+2] "Thanks for the reviews :)" [machinelearning/liftwing/inference-services] - 10https://gerrit.wikimedia.org/r/1023532 (https://phabricator.wikimedia.org/T363294) (owner: 10Kevin Bazira) [12:40:24] (03Merged) 10jenkins-bot: Makefile: add support for logo-detection [machinelearning/liftwing/inference-services] - 10https://gerrit.wikimedia.org/r/1023532 (https://phabricator.wikimedia.org/T363294) (owner: 10Kevin Bazira) [12:48:04] I left instructions on what I tried here -> https://phabricator.wikimedia.org/P61209#246342 since the patch is merged if you want to try it you can skip the part where I checkout the patch [12:50:34] thanks isaranto, I'll investigate further [12:51:39] Thank you kevinbazira but I'd suggest to skip it for now since we have bigger fish to fry :) . Just left it there as reference [12:59:12] (03PS1) 10Klausman: Makefile: cleanup and slight reorganization [machinelearning/liftwing/inference-services] - 10https://gerrit.wikimedia.org/r/1024396 [13:21:00] Morning all [13:29:25] o/ Chris! [13:34:12] green tea on my keyboard 🤦‍♂️ [13:45:46] good morning [13:46:49] (03CR) 10Kevin Bazira: "Thank you for working on this, Tobias! I've left a small note about repetition. The rest LGTM." [machinelearning/liftwing/inference-services] - 10https://gerrit.wikimedia.org/r/1024396 (owner: 10Klausman) [13:49:33] Hey Mercelis! [13:50:34] (03PS2) 10Klausman: Makefile: cleanup and slight reorganization [machinelearning/liftwing/inference-services] - 10https://gerrit.wikimedia.org/r/1024396 [13:50:54] (03CR) 10Klausman: Makefile: cleanup and slight reorganization (031 comment) [machinelearning/liftwing/inference-services] - 10https://gerrit.wikimedia.org/r/1024396 (owner: 10Klausman) [13:51:43] (03CR) 10Kevin Bazira: [C:03+1] Makefile: cleanup and slight reorganization [machinelearning/liftwing/inference-services] - 10https://gerrit.wikimedia.org/r/1024396 (owner: 10Klausman) [13:58:30] 06Machine-Learning-Team, 10Add-Link, 06Growth-Team, 07User-notice: Deploy "add a link" to 18th round of wikis (en.wp and de.wp) - https://phabricator.wikimedia.org/T308144#9744849 (10Trizek-WMF) {T354395} is under progress, we can work on this task so that the configuration option will be available. [14:38:06] klausman: I've rolled out the latest Java 11 security updates on the ml-cache hosts, could you please take care of the Cassandra restarts? [14:38:17] Will do! [14:38:20] cheers [14:38:29] You got a ticket I should update, or just do it? [14:39:38] just do it, for these day to day software updates we usually don't open separate tickets [14:40:01] Ack. I'll reboot them today. They don't serve prod traffic, so I can be quick about it [15:24:55] ack. reboot won't hurt, but strictly speaking we only need a rolling restart of cassandra [15:30:48] all done [15:45:20] cheers [16:07:08] (03CR) 10Klausman: [C:03+2] Makefile: cleanup and slight reorganization [machinelearning/liftwing/inference-services] - 10https://gerrit.wikimedia.org/r/1024396 (owner: 10Klausman) [16:08:44] (03CR) 10Klausman: [V:03+2 C:03+2] Makefile: cleanup and slight reorganization [machinelearning/liftwing/inference-services] - 10https://gerrit.wikimedia.org/r/1024396 (owner: 10Klausman) [16:29:25] (03PS1) 10Ilias Sarantopoulos: utils: slow function execution wrapper [machinelearning/liftwing/inference-services] - 10https://gerrit.wikimedia.org/r/1024425 (https://phabricator.wikimedia.org/T362663) [16:29:54] the patch above --^ is just WIP [16:30:14] going afk folks, have a nice evening! [16:55:54] Now with that excitement out of the way, I'm calling it a day. Seeya tomorrow! [20:06:22] 06Machine-Learning-Team, 13Patch-For-Review: Unsupported lang error for some wiki for revertrisk-language-agnostic calls - https://phabricator.wikimedia.org/T363203#9746292 (10CodeReviewBot) mnz merged https://gitlab.wikimedia.org/repos/research/knowledge_integrity/-/merge_requests/37 feat(featureset): fall b...