[07:19:15] <isaranto>	 Good morning folks!
[07:38:34] <wikibugs>	 (03CR) 10Kevin Bazira: [C: 03+1] "Nice!  LGTM based on the docs: https://huggingface.co/docs/transformers/main/en/quantization#autogptq" [machinelearning/liftwing/inference-services] - 10https://gerrit.wikimedia.org/r/1003488 (https://phabricator.wikimedia.org/T354870) (owner: 10Ilias Sarantopoulos)
[09:56:49] <aiko>	 morning!
[10:00:29] <elukey>	 aiko: o/ I've read about some interesting explainability tools
[10:00:31] <elukey>	 https://github.com/jessevig/bertvi
[10:00:41] <elukey>	 https://github.com/shap/shap
[10:00:59] <elukey>	 https://github.com/marcotcr/lime
[10:01:10] <elukey>	 https://github.com/pytorch/captum
[10:01:41] <elukey>	 (all in https://www.turingpost.com/p/explainableai but I have the full article and I don't know how to share it)
[10:05:13] <isaranto>	 Morning aiko and elukey !
[10:06:11] <isaranto>	 thanks for the links Luca! 
[10:06:16] <klausman>	 Morning!
[10:06:38] <isaranto>	 o/ Tobias
[10:06:39] <aiko>	 hiii elukey :D thanks for sharing. going to read them! 
[10:07:06] <aiko>	 hi Ilias and Tobias o/
[10:07:51] <isaranto>	 from the above I vouch for shap as it heavily used (and mature)
[10:08:05] <klausman>	 Following Chris's question/suggestion on Slack, I switched my Cass POC to using RRLA (instead of ML). It was quite easy, but highlighted a few code smells (hardcoded stuff, mostly).
[10:08:07] <isaranto>	 TIL about the others
[11:12:36] * klausman lunch
[11:41:06] <isaranto>	 ditto
[12:44:45] <isaranto>	 I'm still testing the above patch (quantization on GPU) locally.  Will either update the patch or merge it once done
[12:45:46] <isaranto>	 images take a while to build, one of the following days I'll prioritize building a base image with pytorch as we discussed yesterday
[13:54:21] <wikibugs>	 10Machine-Learning-Team, 10ORES, 10Technical-Debt: Replace usage of wfGetDB() in ORES before the 1.42 cut so it can be hard-deprecated - https://phabricator.wikimedia.org/T357654 (10Jdforrester-WMF)
[14:09:52] <chrisalbon>	 Good morning all
[14:10:57] <chrisalbon>	 Morning elukey
[14:14:38] <klausman>	 Heyo Chris
[14:48:09] <wikibugs>	 (03CR) 10Kevin Bazira: [C: 03+1] "Great! Starting with validating the idea sounds good." [machinelearning/liftwing/inference-services] - 10https://gerrit.wikimedia.org/r/1003488 (https://phabricator.wikimedia.org/T354870) (owner: 10Ilias Sarantopoulos)
[15:34:44] <isaranto>	 Hey Chris o/
[15:36:52] <chrisalbon>	 hello
[15:39:47] <klausman>	 heads up: ml-staging2001 (the only GPU machine in staging will have a super short network flip-flop around the hour). I will drain it beforehand and undrain right after.
[15:47:52] <isaranto>	 ack!
[16:07:42] <klausman>	 and we're back
[17:11:38] <isaranto>	 going afk for the evening folks, cu tomorrow o/
[17:16:01] <klausman>	 \o
[18:16:34] <chrisalbon>	 night all
[20:01:56] <wikibugs>	 10Machine-Learning-Team, 10Research: Explore using revertrisk language agnostic API in a pre-save context - https://phabricator.wikimedia.org/T356102 (10XiaoXiao-WMF) @kostajh to follow up on this - the ML team do not intend to do any work on this, and from our side the language agnostic model is capable to be...
[20:53:29] <wikibugs>	 10Machine-Learning-Team, 10Research: Explore using revertrisk language agnostic API in a pre-save context - https://phabricator.wikimedia.org/T356102 (10kostajh) >>! In T356102#9548565, @XiaoXiao-WMF wrote: > [...] from our side the language agnostic model is capable to be invoked in either situation.  To clar...
[22:47:34] <wikibugs>	 10Machine-Learning-Team, 10Research: Explore using revertrisk language agnostic API in a pre-save context - https://phabricator.wikimedia.org/T356102 (10XiaoXiao-WMF) thanks for the clarification @kostajh. It sounds to me that this is to be discussed with the ML team - it is more about when to invoke the rever...