IRC logs of #shogun for Wednesday, 2017-07-05

--- Log opened Wed Jul 05 00:00:42 2017
@sukey	[https://github.com/shogun-toolbox/shogun] Pull Request https://github.com/shogun-toolbox/shogun/pull/3901 opened by micmn	00:43
-!- mikeling [uid89706@gateway/web/irccloud.com/x-qlobivsojkbftbxq] has joined #shogun		04:04
@wiking	mikeling, how's things?	05:33
mikeling	wiking: Hey, things going well for now. I'm trying to give a fixed seed for each module and see if everything works	05:34
@wiking	mmmm dont need to waste too much time with the seed	05:35
@wiking	as we are going to	05:35
@wiking	change the whole engine :)	05:35
mikeling	but we need to see all the tests passed before we move next step, right?	05:36
mikeling	a lot of test reply on fixed seed to get same result every time	05:37
@wiking	mmm	05:42
@wiking	yeah but	05:42
@wiking	as you gonna change the engine	05:42
@wiking	you'll either need to change the seed	05:42
@wiking	or change the test	05:42
@wiking	as it's gonna be a total different random engine when we switch to c++11 random	05:43
@wiking	you are now having tests failing when u switched from CMath random?	05:43
mikeling	yes, around 30 unit tests failed. Most of them are related to cross validation and MMD	05:45
mikeling	Because they are rely on CMath::randn_double and CMath::normal_random	05:46
mikeling	usually we call CMath_init_random(seed) to give them a fixed seed	05:47
mikeling	after random related functions of CMath gone, we need set seed for they one by one	05:47
@wiking	what do you mean one by one?	05:48
@wiking	ah you mean if you call Obj1 that has	05:48
@wiking	a list of Objects	05:48
@wiking	then if you set Obj1->set_seed()	05:48
@wiking	it does not set the seed for the rest	05:48
@wiking	of the objects of Obj1 right?	05:49
mikeling	wiking: yeah, so here is the thing i do in NeuralNetwork for set seed https://gist.github.com/MikeLing/503e4fe7b09f7f264ce0e30342ece688	05:49
mikeling	when I'm going to set seed for NeuralNetwork	05:49
mikeling	I also need to give a seed for those Neural layers	05:50
@wiking	mmm	05:50
@wiking	yeah actually	05:50
@wiking	we need to think about this a bit	05:50
mikeling	otherwise, the tests may failed, because the Neuralayers will been init randomly	05:50
@wiking	because yeah if you do this per class it's a bit cumbersom	05:51
@wiking	*cumbersome	05:51
mikeling	yes, I know	05:51
mikeling	I also ask Heikos about this in https://github.com/shogun-toolbox/shogun/pull/3888#discussion_r125507162	05:52
@wiking	yeah i saw that one	05:52
mikeling	I'm not sure if it's necessary to set a fixed seed one by one	05:52
@wiking	mikeling, i think	05:53
mikeling	but the unit test will failed if not	05:53
@wiking	honestly	05:53
@wiking	we could have a global static seed	05:53
@wiking	that would solve every problem	05:53
@wiking	right?	05:53
@wiking	mikeling, so the seed could be	05:53
mikeling	yes...but the issue is about to remove the global seed	05:53
mikeling	ok	05:54
@wiking	static uint32_t CSGObject::rng_seed	05:54
@wiking	that is a global variable	05:54
@wiking	that one can set with a static function	05:54
@wiking	of course we would still have	05:54
@wiking	m_rng for every object	05:54
@wiking	but the seed would be shared among them	05:54
mikeling	mmmm, how? I mean we only need this global seed in unit test to make sure all the modules have same seed, right?	06:00
mikeling	So how to use that global seed(how to give it to the modules we used	06:01
@wiking	nono	06:04
@wiking	i mean you might wanna have the set seed thing to be possible	06:05
@wiking	in runtime as well	06:05
@wiking	it's easy	06:05
@wiking	mikeling, https://gist.github.com/vigsterkr/3268e9482b90250b6450f6187e647cd7	06:06
@wiking	and when you create the random object	06:06
@wiking	you pass CSGObject::m_seed to CRandome	06:07
@wiking	that's all	06:07
mikeling	wiking: I see. I will try it right away. Thank you!	06:10
-shogun-buildbot:#shogun- Build nightly_all #8 is complete: Success [build successful] - http://buildbot.shogun-toolbox.org:8080/#builders/22/builds/8		06:33
@wiking	Trixis, i have a jar for you if you care to tests	07:08
@wiking	lisitsyn, ping	08:23
lisitsyn	wiking: hey	08:23
@wiking	lisitsyn, ok so how do you feel about static/global prng seed?	08:23
lisitsyn	global should go :)	08:24
lisitsyn	no?	08:24
@wiking	i mean it's the same :)	08:25
@wiking	static CSGObject::seed	08:25
lisitsyn	uh why?	08:25
@wiking	because we have the story of	08:26
@wiking	MMD and NN	08:26
@wiking	SGobject->[SGObject]+	08:26
@wiking	and then how do you set the seed the same	08:26
lisitsyn	aha	08:26
@wiking	among those object	08:26
@wiking	?	08:26
@wiking	hence the idea of a static seed	08:27
@wiking	i mean honestly	08:27
@wiking	if we go with this	08:27
@wiking	we can just drop the SGObjcet::Random	08:27
@wiking	as it's just a hazard thing	08:27
@wiking	to assume	08:27
@wiking	to have a stateful random	08:27
@wiking	you should always generate a PRNG when you need it	08:27
@wiking	with an initial state	08:27
@wiking	see what i mean?	08:27
lisitsyn	yes	08:27
@wiking	so just lets have some sort of a global seed	08:28
@wiking	that is possible to set	08:28
@wiking	otherwise the seed is set from /dev/urandom	08:28
@wiking	and that's all	08:28
lisitsyn	but static is global right?	08:28
lisitsyn	what's changed then :)	08:28
@wiking	that the PRNG is not global	08:28
@wiking	because till now	08:29
@wiking	we had a global prng	08:29
@wiking	that was used by everybody in the same process space	08:29
@wiking	shitty scenario: 2 threads/forks with each having 1 model that uses the prng	08:29
@wiking	even if you set the seed	08:30
@wiking	you cannot predict the PRNG's order to each model	08:30
@wiking	as they share the PRNG	08:30
@wiking	so we definitely wanna clear out to have global prng	08:30
lisitsyn	but don't they share static?	08:30
@wiking	they would share the starting state	08:30
@wiking	the seed	08:30
@wiking	if you set it	08:30
@wiking	see for example	08:31
@wiking	https://github.com/shogun-toolbox/shogun/blob/develop/src/shogun/mathematics/Random.cpp#L346-L366	08:31
@wiking	this is what by default the seed of the PRNG	08:31
@wiking	which should be pretty random	08:32
@wiking	:)	08:32
@wiking	but if one sets the seed	08:32
lisitsyn	don't you want thread-local static?	08:32
@wiking	mmm	08:32
@wiking	yeah we could do that	08:33
@wiking	but yeah we need to have something	08:33
@wiking	that is shared	08:33
@wiking	among some objects	08:33
@wiking	otherwise	08:33
lisitsyn	because simple static is shared just like the global	08:33
@wiking	lisitsyn, currenly i'm looking at srandom() random() story of uni	08:33
@wiking	*unix	08:33
@wiking	srandom sets the whole state of random	08:33
@wiking	process space wise	08:34
@wiking	afaik	08:34
lisitsyn	can we make it non-static but make the default value static?	08:35
@wiking	eh?	08:35
lisitsyn	like if you set the seed that's what you gonna use	08:35
@wiking	i'm not so sure if i follow	08:35
@wiking	yeye i mean of course	08:35
@wiking	the seed would by default be random	08:36
@wiking	unless u set it's value	08:36
lisitsyn	so global/static is the default value for new objects	08:36
@wiking	ah and for some reason you wanna support	08:36
@wiking	the seet	08:36
@wiking	*seed	08:36
@wiking	per object?	08:36
lisitsyn	so obj->seed and obj->default_seed	08:36
lisitsyn	yeah	08:36
@wiking	use-case?	08:36
@wiking	for the object level seed?	08:37
lisitsyn	training the same neural network with different initializations	08:37
@wiking	yeah how do you make that the NN has all the seeds set correctly	08:37
@wiking	:)	08:37
lisitsyn	we need a way to inherit such things somehow	08:38
@wiking	yeah that's what mikeling started	08:38
@wiking	with a custom solution	08:39
@wiking	i.e. if you are the developer of the algo	08:39
@wiking	you make sure that the set_seed set's all the objects you hold have their seed set	08:39
@wiking	honestly currenlty	08:39
@wiking	i dont see why we wanna have	08:39
@wiking	a SGObject::CRandom	08:39
@wiking	i mean that each and every SGObject has it's prng	08:40
@wiking	those prng should be created somewhere under the hood of a function	08:40
lisitsyn	no but seeds should be inherited	08:40
lisitsyn	it is a state but a bit better :D	08:41
-!- mikeling is now known as mikeling\|afk		08:50
lisitsyn	wiking: well in the end we can refer to sklearn	08:50
lisitsyn	:D	08:50
lisitsyn	they provide random_state arg	08:50
lisitsyn	per run	08:50
@wiking	lisitsyn, so yeah	08:50
@wiking	they have a per object seed	08:51
@wiking	but of course	08:51
@wiking	i guess because of the 'design' of python	08:51
@wiking	the whole algo is written in one class	08:51
@wiking	probably even in one function :P	08:51
lisitsyn	yeah but we are in the same boat	08:51
@wiking	not really	08:51
lisitsyn	we're supposed to be used from python	08:51
@wiking	we have objects of objects	08:51
@wiking	i understand that	08:52
@wiking	but under the hood	08:52
@wiking	you have objects that might wanna use a seed	08:52
@wiking	so thread_local ?	08:53
@wiking	or let the algo developer	08:54
@wiking	decide and implement this part	08:54
@wiking	?	08:54
@wiking	Trixis, ping me when u r around	08:58
@wiking	lisitsyn, so what do we do ? :)	09:05
lisitsyn	wiking: yolo	09:06
lisitsyn	wiking: seed is thread-local and initialized with something random unless global seed is set up	09:07
@wiking	global?	09:07
@wiking	:D	09:07
lisitsyn	static whatever	09:07
@wiking	i dont get it now	09:07
lisitsyn	I see no difference between static and global :P	09:07
@wiking	mmm	09:07
@wiking	yeah there's none	09:08
@wiking	but the question is	09:08
@wiking	do u want that thing to be	09:08
@wiking	thread_local?	09:08
@wiking	or global?	09:08
@wiking	the static	09:08
lisitsyn	initial aka default seed or current seed?	09:08
@wiking	so fuc kthis	09:09
@wiking	:)	09:09
@wiking	i dont understand anything	09:09
@wiking	why do u need a separate seed/	09:09
lisitsyn	:D	09:09
@wiking	why dont we ahve	09:09
@wiking	either a thread local	09:09
@wiking	or full global seed	09:09
@wiking	that is by default random	09:09
@wiking	but if you set it	09:09
@wiking	then it's set	09:09
@wiking	that's it	09:09
lisitsyn	yeah might be fine as well	09:09
@wiking	and then either objects that are in the same thread are using the same seed	09:10
@wiking	or the every object in the whole process space	09:10
@wiking	the only question now that if we use	09:10
@wiking	thread_local	09:10
@wiking	we gonna have a hard time with openmp stuff	09:10
@wiking	:D	09:10
lisitsyn	then no thread-local I guess	09:12
-!- geektoni [~geektoni@93-34-128-38.ip49.fastwebnet.it] has joined #shogun		09:14
@wiking	so global	09:14
@wiking	:)	09:14
@wiking	mikeling\|afk, ^ we gonna have one global seed variable	09:14
-!- iglesiasg [~iglesiasg@217.119.234.214] has joined #shogun		09:27
-!- mode/#shogun [+o iglesiasg] by ChanServ		09:27
@wiking	lisitsyn, in feature/ml-shogun-java switched things to ml.shogun in java	09:45
lisitsyn	wiking: RENAME ALL THE SHIT	09:47
lisitsyn	YOLO	09:47
lisitsyn	:D	09:47
lisitsyn	wiking: https://github.com/shogun-toolbox/shogun/pull/3898 YOLO?	09:47
lisitsyn	wiking: yolo push da button	09:56
lisitsyn	:P	09:56
@wiking	lisitsyn, u push	09:57
@wiking	:)	09:57
@sukey	[https://github.com/shogun-toolbox/shogun] Pull Request https://github.com/shogun-toolbox/shogun/pull/3898 merged by lisitsyn	09:58
@sukey	[https://github.com/shogun-toolbox/shogun] lisitsyn pushed 3 commits:	09:58
@sukey	https://github.com/shogun-toolbox/shogun/commit/75657cbe565c492e4963104d5d980acd1d82416b	09:58
@sukey	https://github.com/shogun-toolbox/shogun/commit/ea53b8b0638bd06fb1689d3379a0fa8bfd10a2a4	09:58
lisitsyn	yolo!!!!!!!	09:58
@sukey	https://github.com/shogun-toolbox/shogun/commit/6d9464ee3dfc576c25060f2654822e208a7ef5c3	09:58
@wiking	lets see the buildbots	09:59
-shogun-buildbot:#shogun- Build deb4 - python3 #45 is complete: Failure [failed python (failure)] - http://buildbot.shogun-toolbox.org:8080/#builders/28/builds/45		10:01
lisitsyn	YYYY	10:03
-shogun-buildbot:#shogun- Build deb3 - interfaces #14 is complete: Failure [failed python (failure)] - http://buildbot.shogun-toolbox.org:8080/#builders/37/builds/14		10:08
@wiking	:)	10:09
@wiking	fail fail	10:09
lisitsyn	lol	10:09
lisitsyn	ok I had no doxygen locally	10:09
@wiking	lisitsyn, yeah there's a dir missing there still	10:09
lisitsyn	must be easy fix	10:10
@wiking	lisitsyn, Tags fix?	10:18
lisitsyn	wiking: tags fix?	10:18
@sukey	[https://github.com/shogun-toolbox/shogun] Pull Request https://github.com/shogun-toolbox/shogun/pull/3877 synchronized by geektoni	10:18
lisitsyn	que?	10:18
lisitsyn	señor?	10:19
@wiking	lisitsyn, geektoni needs to start switching to	10:19
geektoni	lisitsyn: non-owning any I think ;)	10:19
lisitsyn	no comprende lo siento	10:19
lisitsyn	no hablo ingles	10:19
@wiking	lisitsyn, :D	10:20
lisitsyn	wiking: I added this thing to switch	10:21
lisitsyn	non-owning is next once I finish my procrastination renaming	10:21
@wiking	geektoni, mmm did u remove the default fitlter of the observable?	10:21
-!- mikeling\|afk is now known as mikeling		10:22
@wiking	mikeling, welcomeback :)	10:22
geektoni	wiking: default filter? :P	10:22
@wiking	geektoni, default = no filter of the params	10:22
mikeling	wiking: :D	10:22
geektoni	ahh yes yes	10:23
@wiking	mikeling, so the idea is that there's a static seed that is by default inited like this = https://github.com/shogun-toolbox/shogun/blob/develop/src/shogun/mathematics/Random.cpp#L346-L367	10:23
geektoni	wiking: if ParameterObserverScalar is created without providing a parameter list	10:23
@wiking	and every CRandom uses this seed	10:23
@wiking	geektoni, \o/	10:24
@wiking	mikeling, dont you have some SEO friends in china who could get shogun into a better position on google search terms?	10:25
@wiking	:D	10:25
lisitsyn	oh that's so racist like all chinese are doing SEO :D :D :D	10:25
lisitsyn	SCNR	10:26
mikeling	wiking: no, but i guess i could find one :D	10:26
@wiking	:D	10:26
@wiking	lisitsyn, no they have enough resources to do backlinks :)	10:26
lisitsyn	let me show you the real racism	10:26
lisitsyn	mikeling: could you give me contacts of your friends running a mining farm	10:27
@wiking	:D	10:27
mikeling	lisitsyn: mining farm? you mean the bitcoin?	10:27
mikeling	:)	10:27
@wiking	mikeling, or ethereum	10:27
@wiking	:)	10:27
mikeling	I wise I know someone who run a mining farm, so I can borrow a gtx 1080 from them	10:29
@wiking	:>	10:30
@wiking	mikeling, rx480 or rx580 is better :)	10:30
mikeling	:D :D :D	10:30
@wiking	although it's impossible to find one	10:30
@wiking	:D	10:30
lisitsyn	I wasn't bothered with eco before	10:31
@wiking	get an ico :)	10:32
@wiking	i've got tokens in this one	10:32
lisitsyn	but all these guys are going to produce more CO2 than all the cars	10:32
@wiking	https://polybius.io/	10:32
@wiking	they already raised 29M usd	10:32
@wiking	:D	10:32
lisitsyn	ohh that's so bullshit bingo	10:33
lisitsyn	:D	10:33
@wiking	:D	10:34
lisitsyn	IOT BLOCKCHAIN OMG	10:34
@wiking	yep	10:34
@wiking	all this is crazy	10:34
@wiking	but the banking license is possible :D	10:34
@wiking	if of all things	10:34
@wiking	at least that goes through	10:34
@wiking	that's great	10:34
@wiking	lisitsyn, btw https://gist.github.com/geektoni/c3e2b3802ca5910737c6dd68665c1fb5	10:38
@wiking	lisitsyn, pip install tensorboard	10:38
@wiking	download the binary file and then	10:38
@wiking	tensorboard --logdir <downloaded dir>	10:38
lisitsyn	do you know a way to merge them into one grap?	10:41
lisitsyn	would be fun to see something like lasso path	10:41
geektoni	lisitsyn: you can use a regex like weights_* to move all the graphs into one Tensorboard tab	10:45
lisitsyn	yeah but they are still separate graphs?	10:45
lisitsyn	geektoni: I am sorry to change my mind but probably exporting to 'histograms' would probably be better	10:45
lisitsyn	:D	10:46
@wiking	lisitsyn, lol motherfucker	10:46
@wiking	but yeah it's just a different type of observable	10:46
geektoni	wiking: we have just to change the TBOutputFormat actually	10:47
@wiking	?	10:47
@wiking	you mean event	10:47
@wiking	in the event instead of scalar summary	10:47
geektoni	yes, I mean we need only to adding a new method that returns an event with an Histogram inside	10:47
geektoni	exactly	10:47
@wiking	geektoni, but that is not on the tflogger level	10:48
@wiking	but in shogun	10:48
@wiking	right?	10:48
lisitsyn	it is especially useful to see histogram of weights in neural nets	10:48
@wiking	i mean i would hope so :)	10:48
geektoni	wiking: TFLogger won't be touched	10:48
geektoni	it is all on Shogun's side	10:48
@wiking	yep	10:48
@wiking	well i guess you need to add a different observable no?	10:48
@wiking	now it's *Scalar	10:49
@wiking	but you would have a *Histogram?	10:49
lisitsyn	I think it computes histograms for you	10:49
@wiking	who?	10:49
lisitsyn	tensorboard	10:49
lisitsyn	iirc in keras you just set up exporting weights	10:50
@wiking	lisitsyn, mmm no you have to dump it in a different format	10:50
@wiking	https://max.io/bash.html	10:50
lisitsyn	okie	10:50
@wiking	lisitsyn, TB uses TF's Event object	10:50
lisitsyn	no idea	10:50
lisitsyn	:D	10:50
@wiking	and it has different type of Events	10:50
@wiking	one is just a scalar	10:50
@wiking	other is histogram	10:50
@wiking	there's img	10:50
@wiking	etc	10:50
@wiking	but yeah it's possible to create a histogram	10:51
lisitsyn	wiking: can you stream videos?	10:51
lisitsyn	streaming some pornhub while learning some model would be a killer feature	10:52
@wiking	lisitsyn, u can stream audio and img	10:53
@wiking	img stream -> video	10:54
@wiking	:D	10:54
lisitsyn	oh with audio it is even better	10:54
geektoni	wiking: I think we would need to manually populate the Histogram object. Anyway, Histogram are generated from a series of vectors, so we need also to add a SGObject method called observe_vector(std::vector) which will emit an Any wrapped vector.	10:54
@wiking	mmm yeah	10:55
@wiking	i mean	10:55
@wiking	currently	10:55
@wiking	we said that the observation of params are up to the algo developer right?	10:55
@wiking	so how you actually populate a Histogram, you manually set the histogram elements one-by-one (populate a vector and set that to be the histogram) or use an actual vector to be monitored	10:56
@wiking	it's a different story	10:56
@wiking	but i guess geektoni the weights_* are coming from a float64_t* no?	10:57
geektoni	no, they come from single emitted object	10:57
@wiking	yeah but	10:57
@wiking	under the hood in lasso	10:57
@wiking	isn't those weights stored as a float64_t*?	10:58
geektoni	yes, they are.	10:58
@wiking	so yeah that's a std::vector in a way :P	10:58
geektoni	yes it is. We need only to emit the entire vector and build the Histogram from there :P	11:00
@wiking	\o//	11:02
geektoni	wiking: https://github.com/tensorflow/tensorflow/tree/master/tensorflow/core/lib/histogram	11:05
geektoni	it's just a matter of copy-paste ;)	11:05
@wiking	btw what's gtl	11:06
@wiking	namespace?	11:06
@wiking	i guess https://github.com/tensorflow/tensorflow/tree/master/tensorflow/core/lib/gtl	11:07
geektoni	seems to be a wrapper for std::vector or similar structures	11:08
lisitsyn	I think it is like stl	11:15
lisitsyn	but gtl	11:15
lisitsyn	:)	11:15
lisitsyn	like parts of google's standard library	11:15
lisitsyn	they have a tool called MOE	11:16
lisitsyn	they use it to export their mono repo	11:16
lisitsyn	into opensource projects	11:16
@sukey	[https://github.com/shogun-toolbox/shogun] New commit https://github.com/shogun-toolbox/shogun/commit/523941577baebec6a3c26bffa61fe64d39faa758 by lisitsyn	11:37
lisitsyn	wiking: oh new buildbot is like	11:39
lisitsyn	oh there is new bootstrap	11:39
lisitsyn	lets use all the elements	11:40
lisitsyn	wiking: I fixed the doxygen failure	11:49
-!- HeikoS [~heiko@host-92-0-169-11.as43234.net] has joined #shogun		11:50
-!- mode/#shogun [+o HeikoS] by ChanServ		11:50
-shogun-buildbot:#shogun- Build deb4 - python3 #48 is complete: Success [build successful] - http://buildbot.shogun-toolbox.org:8080/#builders/28/builds/48		11:51
@wiking	goodie	11:52
@wiking	lisitsyn, here?	11:59
lisitsyn	wiking: yeas	11:59
@wiking	ok so	11:59
@wiking	have you looked into the descriptor of CoreML	11:59
@wiking	(the protobuf files)	11:59
lisitsyn	yes	11:59
@wiking	ok so in Model you realised that there's an actual	12:00
@wiking	Pipeline	12:00
@wiking	right	12:00
@wiking	?	12:00
@wiking	which is more or less a Model*	12:00
@wiking	so i was wondering	12:00
@wiking	how about if we would actually use	12:00
@wiking	within shogun to define pipelines?	12:01
@wiking	so that we dont only have a shogun model stack -> coreml	12:01
@wiking	but as well we would use that descriptor to define pipelines in shogun	12:01
@wiking	and train it	12:01
@wiking	btw lisitsyn https://numer.ai/whitepaper.pdf	12:02
@wiking	:D	12:02
lisitsyn	haha	12:03
@wiking	geektoni micmn none of you is interested to write a shogun pipeline for numer.ai? :D	12:03
lisitsyn	wiking: I agree we should fit the models like PMML and CoreML	12:04
lisitsyn	like change structure a bit	12:04
lisitsyn	so it is more conventional	12:04
@wiking	yep	12:04
@wiking	but what's your input on that pipeline story/	12:04
micmn	what? :)	12:04
@wiking	the reason i like the coreml	12:04
@wiking	that it's all nicely fixed typed and it's language agnostic :)	12:05
@wiking	micmn, check numer.ai :)	12:05
lisitsyn	wiking: yes I am a big fan of coreml thingy also	12:06
lisitsyn	a bit tired explaining why this matters	12:06
lisitsyn	they do they freaking python fragile shit	12:06
geektoni	wiking: I quote micmn, what? xD	12:06
lisitsyn	and say it is like industry ml stuff	12:07
@wiking	lisitsyn, indeed	12:07
@wiking	so lets get ahead of the game for once :)	12:07
@wiking	but yeah	12:07
@wiking	numer.ai	12:07
@wiking	earn some numerai tokens with ML	12:07
@wiking	:D	12:07
@wiking	basically trading with ML	12:08
@wiking	you get a binary classification problem	12:09
@wiking	and send in the solution	12:09
lisitsyn	sounds something like quantopian?	12:09
@wiking	and your model gets ranked	12:09
@wiking	yeah	12:09
lisitsyn	oh that sounds smart	12:09
@wiking	it's just that in this case	12:10
@wiking	you dont need to handle time series or anything	12:10
@wiking	the problem is mapped to a binary classification	12:10
@wiking	that's all you need to solve	12:10
lisitsyn	ahh so you participate in an auction	12:10
lisitsyn	in which your bid is related to your confidence in overfitting	12:10
@wiking	ye	12:11
@wiking	and in past you've got BTC	12:11
lisitsyn	so you basically lose all the time whne you overfit	12:11
lisitsyn	smart!	12:11
@wiking	but now you get their own token	12:11
@wiking	anyhow that token is tradable	12:11
lisitsyn	based on eth	12:11
lisitsyn	ohahoha	12:11
@wiking	so you can convert that to any cryptocoin	12:11
@wiking	so it's ok	12:11
@wiking	anyhow	12:11
@wiking	would be great if there would be some demo of shogun	12:11
@wiking	being used for numerai	12:11
@wiking	which basically	12:12
@wiking	means that you need to have a pipeline :)))	12:12
lisitsyn	wiking: ok step one	12:14
lisitsyn	do we put coreml protos directly into	12:14
lisitsyn	the repo oder?	12:14
@wiking	mmm	12:14
@wiking	good question	12:14
@wiking	actually do we need to? :)	12:15
lisitsyn	well we should have them somewhere	12:15
@wiking	mmm yeah true	12:15
@wiking	i wonder if there's a repo for this already on github	12:15
lisitsyn	i don't see anything reliable	12:16
lisitsyn	apple can't into opensource bhaha	12:16
@wiking	:>>>.	12:16
lisitsyn	lets just put them	12:17
lisitsyn	if license	12:17
lisitsyn	wiking: BSD-3	12:17
lisitsyn	let me PR	12:18
@wiking	:)	12:18
@wiking	lisitsyn, do we wann have it directly in shogun repo?	12:19
@wiking	maybe a subrepo?	12:19
@wiking	jsut because we'll hve to keep track of changes	12:19
@wiking	or soething	12:19
lisitsyn	wiking: yeah third_party	12:19
@wiking	although i hate	12:19
@wiking	submodules	12:19
lisitsyn	me too	12:19
@wiking	because of git grep	12:19
lisitsyn	I see no need	12:19
@wiking	yeah go with the third_party	12:19
@wiking	buuut	12:20
@wiking	i'm wondering if we wanna have it in the repo	12:20
@wiking	or have like a cmake script?	12:20
lisitsyn	wiking: which does what?	12:20
@wiking	that would always download you the shit?	12:20
@wiking	(see tensorflow: https://github.com/tensorflow/tensorflow/tree/master/third_party	12:20
lisitsyn	no no it would be unstable	12:20
lisitsyn	oh tf really downloads all the stuff	12:21
@wiking	i hate a build	12:21
@wiking	that cannot be done offline ;0	12:21
@wiking	)	12:21
@wiking	:)	12:21
lisitsyn	yes	12:22
lisitsyn	so lets put them	12:22
lisitsyn	where?	12:22
lisitsyn	I can't put them to third_party because it is a generated directory	12:22
@wiking	mm	12:22
@wiking	we can change the generated dir's name	12:23
-shogun-buildbot:#shogun- Build deb3 - interfaces #17 is complete: Success [build successful] - http://buildbot.shogun-toolbox.org:8080/#builders/37/builds/17		12:23
lisitsyn	wiking: uhm can't we say	12:24
lisitsyn	coreml is an interface	12:24
@wiking	:D	12:24
lisitsyn	src/interface/coreml	12:24
lisitsyn	:)	12:24
@wiking	sure	12:24
@wiking	sounds like a good stuff :)))	12:24
lisitsyn	why not	12:24
@wiking	although it'll bring in some confusion	12:24
@wiking	but who cares	12:24
@wiking	let's just do it	12:24
lisitsyn	src/interface/pmml	12:24
@wiking	pma?	12:24
lisitsyn	PMML	12:25
lisitsyn	:)	12:25
@wiking	i mean there's a new thingy	12:25
@wiking	PFA	12:25
lisitsyn	ah	12:25
lisitsyn	yes all the shit around	12:25
lisitsyn	PFA PMML IOT YOLO	12:25
@wiking	http://dmg.org/pfa/docs/tutorial3/	12:26
lisitsyn	interface/bitcoin	12:26
@wiking	yes yes	12:26
@wiking	all of that	12:26
lisitsyn	interface/eth	12:26
@wiking	but yeah put in coreml	12:26
lisitsyn	interface/shitcoin	12:26
@wiking	but you know what's the problem	12:26
@sukey	[https://github.com/shogun-toolbox/shogun] Pull Request https://github.com/shogun-toolbox/shogun/pull/3902 opened by lisitsyn	12:26
lisitsyn	yolo	12:26
@wiking	that we need to precompile	12:26
@wiking	the fucking coreml	12:26
@wiking	protobuf files	12:27
@wiking	right?	12:27
lisitsyn	yes	12:27
lisitsyn	I'd add a cmakelists	12:27
lisitsyn	for that	12:27
lisitsyn	later	12:27
@wiking	yeye it's easy	12:27
@wiking	and then all shogun header's would start depending	12:27
@wiking	on coreml?	12:27
lisitsyn	lets see if Herr HeikoS loves it	12:27
lisitsyn	no why?	12:27
lisitsyn	vice versa	12:27
@wiking	i mean how do you dump	12:27
@wiking	a shogun model	12:28
@wiking	into coreml format?	12:28
lisitsyn	we'd have a code that transforms it back and forth maybe	12:28
@wiking	but then this whole shit	12:28
@HeikoS	lisitsyn: ?	12:28
lisitsyn	goes to src	12:28
@wiking	could be an external library no? :D	12:28
@wiking	which depends on shogun and coreml	12:28
@wiking	:)))	12:28
lisitsyn	HeikoS: https://github.com/shogun-toolbox/shogun/pull/3902	12:28
lisitsyn	wiking: yeah just like anything else in shogun :P	12:29
@wiking	lisitsyn, we are brewing this repo to be a monster	12:29
@wiking	:D	12:29
lisitsyn	yeah why not	12:29
@wiking	okok i know linux kernel is one repo as well	12:29
@wiking	:D	12:29
lisitsyn	coreml is important	12:29
lisitsyn	wiking: GOOGLE is one repo	12:29
@HeikoS	lisitsyn: is that copy paste or did you modify things?	12:29
lisitsyn	HeikoS: copy paste	12:29
lisitsyn	download, cp -r	12:29
lisitsyn	:)	12:29
@HeikoS	you gotta give me some context here	12:29
lisitsyn	HeikoS: oh apple released a set of proto definitions for ML models	12:30
@wiking	HeikoS, https://developer.apple.com/documentation/coreml	12:30
lisitsyn	if you follow these protos	12:30
lisitsyn	you can load your model in iOS 11	12:30
lisitsyn	and next macosx	12:30
@HeikoS	I see	12:30
lisitsyn	and apply it immediately	12:30
lisitsyn	without any code	12:30
@HeikoS	cool feature	12:30
lisitsyn	just load('serialize_model_in_proto_format')	12:30
@wiking	they released a python lib	12:30
@HeikoS	yeah I see	12:30
@HeikoS	cool stuff	12:30
@HeikoS	I though might question priority a bit ;)	12:30
@wiking	that can convert an xgboost, scikitlearn and whatnot to this format	12:30
@wiking	of course not shogun	12:31
@wiking	:)	12:31
lisitsyn	it is of highest priority	12:31
lisitsyn	:)	12:31
@wiking	i would say it's really important	12:31
@wiking	:D	12:31
lisitsyn	really	12:31
@HeikoS	it is	12:31
@HeikoS	but I guess we have a list of very important things haha	12:31
@HeikoS	which is epic	12:31
@HeikoS	no I mean do it	12:31
@HeikoS	DO IT	12:31
lisitsyn	yes but I can't follow the list	12:31
lisitsyn	I am too young to live by rules	12:31
lisitsyn	:P	12:31
@HeikoS	lisitsyn: the only thing that I would try to avoid	12:32
@HeikoS	is starting this, and stopping when it is half baked	12:32
@HeikoS	e.g. just importing the definitions and then never touching them again	12:32
lisitsyn	a PR with full support would be impossible	12:32
@HeikoS	sure sure	12:33
lisitsyn	HeikoS: so did you see we're shogun again?	12:33
lisitsyn	make shogun shogun again	12:33
@HeikoS	lisitsyn: yes I love that	12:33
@HeikoS	so good	12:33
lisitsyn	I am going to fix the setup.py	12:33
lisitsyn	it doesnt work for me	12:33
lisitsyn	not my changes it just didn't work	12:33
@wiking	lisitsyn, hahahahah	12:33
@wiking	lisitsyn, setup.py was never working	12:34
lisitsyn	ah	12:34
lisitsyn	I see	12:34
lisitsyn	I suspected that	12:34
@wiking	lisitsyn, and btw if you wanna do anything with that	12:34
lisitsyn	I do	12:34
@wiking	first i would move setup.py to src/interfaces/python	12:34
lisitsyn	I want pip install shogun	12:34
lisitsyn	good idea	12:34
@wiking	yeye i get thatt	12:34
@wiking	lisitsyn, if i were you	12:34
@wiking	lisitsyn, copy-paste tf's pip_install/script stuff	12:34
@wiking	:)	12:34
@wiking	it's pretty much what we need	12:34
lisitsyn	I'll check	12:35
@HeikoS	lisitsyn: there is also the conda stuff that dougal did	12:35
@HeikoS	I shared a link a while ago with wiking	12:35
lisitsyn	I don't really get why conda	12:35
lisitsyn	when pip	12:35
@HeikoS	wiking: btw I was looking for the bsd_license build	12:35
@HeikoS	but buildbot was not responding	12:35
@HeikoS	also not in irc	12:35
@HeikoS	lisitsyn: conda ships MKL for example	12:36
@HeikoS	so numpy is quite a bit fater	12:36
lisitsyn	yeah but installing it with pip into conda	12:36
@HeikoS	and multicore	12:36
lisitsyn	also works	12:36
lisitsyn	no?	12:36
@HeikoS	yes I think	12:36
@wiking	who cares	12:36
@wiking	we need both	12:36
@HeikoS	but that usually creates a mess	12:36
lisitsyn	pip first	12:36
@HeikoS	yep	12:36
@HeikoS	++	12:36
@wiking	HeikoS, what's the problem?	12:36
@HeikoS	shogun-buildbot: force build "nightly_bsd_license"	12:36
@wiking	and?	12:37
lisitsyn	shogun-buildbot: work you fcker	12:37
@wiking	it's doing it	12:37
@wiking	as it is written here http://buildbot.shogun-toolbox.org:8080/#/builders/4	12:37
lisitsyn	haha	12:37
lisitsyn	it is just quiet	12:37
@wiking	shogun-buildbot: dance	12:37
-shogun-buildbot:#shogun- <(^.^<)		12:37
-shogun-buildbot:#shogun- <(^.^)>		12:37
-shogun-buildbot:#shogun- (>^.^)>		12:37
@HeikoS	ah	12:37
-shogun-buildbot:#shogun- (7^.^)7		12:37
@HeikoS	mmg	12:37
-shogun-buildbot:#shogun- (>^.^<)		12:37
@wiking	it's just not communicating too much	12:37
@HeikoS	just doesnt asnwer	12:37
@HeikoS	okok	12:37
@HeikoS	thx	12:37
lisitsyn	a bit autist	12:37
@wiking	yeah	12:37
@HeikoS	yesterday the website kind of froze for me	12:37
@wiking	but it works	12:37
@HeikoS	but maybe that was just hickup	12:37
@wiking	HeikoS, i guess you didn't use	12:37
@wiking	the right addreess :)))	12:37
@HeikoS	the 8080	12:37
@wiking	you must have used http://buildbot.shogun-toolbox.org	12:37
@wiking	instead of http://buildbot.shogun-toolbox.org:8080	12:37
@HeikoS	aaaah	12:37
@HeikoS	yeah ok	12:38
@wiking	i cannot do anything abou tthat	12:38
@HeikoS	no worries	12:38
@HeikoS	just didnt know	12:38
@wiking	7nn has a shitty proxy on :80	12:38
@wiking	HeikoS, read slack ;)	12:38
@HeikoS	eeehm ;)	12:38
@wiking	HeikoS, announced on #general	12:38
@HeikoS	wiking: yesterday a guy came	12:38
@HeikoS	who wanted to link shogun in non gpl fashion	12:39
@wiking	you mean yamz ?	12:39
@HeikoS	yes	12:39
@HeikoS	so I thought I push moving the files a bit	12:39
@wiking	i saw couple of those msgs you head	12:39
@wiking	*had	12:39
@HeikoS	so we make progress there	12:39
@wiking	sure thing the nightly is doing it's job	12:39
@wiking	:)	12:39
@HeikoS	kk	12:39
@wiking	as you can see	12:39
@HeikoS	wiking: question	12:39
@HeikoS	I will not update formatting of the files I move ok?	12:39
@HeikoS	rather do that in batch once all are moved	12:39
@wiking	sure sure	12:39
@HeikoS	so I will merge the PR	12:39
@wiking	go ahrad	12:39
@HeikoS	kk	12:39
@wiking	*ahead	12:40
@wiking	just one thing	12:40
@wiking	!!!	12:40
@sukey	[https://github.com/shogun-toolbox/shogun] Pull Request https://github.com/shogun-toolbox/shogun/pull/3900 merged by karlnapf	12:40
@sukey	[https://github.com/shogun-toolbox/shogun] karlnapf pushed 2 commits:	12:40
@sukey	https://github.com/shogun-toolbox/shogun/commit/d20fdb4d43c30edec1297fde971c0bd4bf0a557e	12:40
@wiking	hahaha	12:40
@wiking	ok	12:40
@sukey	https://github.com/shogun-toolbox/shogun/commit/67e80e63446e45191477d5a783edf59e2eb9e1f4	12:40
@wiking	nevermind	12:40
@HeikoS	aaaaaaaaaah	12:40
@wiking	there was 2 things	12:40
@wiking	could you start using	12:40
@wiking	rebase+merge	12:40
@wiking	instead of merge pull request	12:40
@wiking	in PRs	12:40
@wiking	plz	12:40
@HeikoS	ok	12:40
@wiking	stupid buildbot needs some love	12:40
@wiking	an dnow it builds	12:40
@wiking	every commit	12:40
@wiking	one by one	12:40
@wiking	:)	12:40
@HeikoS	ah I see	12:40
@wiking	second	12:40
@wiking	when you install the gpl-ed version	12:40
@HeikoS	I can also feature branch this	12:40
@wiking	there's no ___ right?	12:40
@wiking	or i mean could you cheack	12:41
@wiking	*check	12:41
@HeikoS	and use travis	12:41
@HeikoS	and then merge at the end	12:41
@wiking	that the headers are correctly installed	12:41
@HeikoS	yeah I can do that	12:41
@wiking	kk	12:41
@wiking	thnx	12:41
@HeikoS	for now I just want to move	12:41
@HeikoS	there is some work needed after all has been moved	12:41
@wiking	sure sure	12:41
@wiking	i'm just saying tha ti'm a bit afraid of that ___ in build	12:41
@HeikoS	yep will look into it	12:41
@HeikoS	shogun-buildbot: force build "nightly_bsd_license"	12:45
micmn	wiking: https://github.com/shogun-toolbox/shogun-data/pull/138 took me a while to setup the meta* machinery	12:48
-!- yamz [400789b6@gateway/web/freenode/ip.64.7.137.182] has quit [Ping timeout: 260 seconds]		12:52
geektoni	wiking: https://gist.github.com/geektoni/c3e2b3802ca5910737c6dd68665c1fb5 we have histograms!	12:54
@HeikoS	micmn: you will have to update the submodule in the main repo as well to make this work	12:55
micmn	yes thx	12:55
@HeikoS	geektoni: that is f**** amazing	12:56
@HeikoS	should be applied to the statistical testing framework by rahul	12:56
@HeikoS	geektoni: btw, this is very cool to apply to x-validation	12:57
@HeikoS	where one is interested in intermediate results of the folds	12:57
@HeikoS	we have some shitty API for that, but this is way better	12:57
-shogun-buildbot:#shogun- Build nightly_bsd_license #11 is complete: Failure [failed compile (failure)] - http://buildbot.shogun-toolbox.org:8080/#builders/4/builds/11		12:58
@sukey	[https://github.com/shogun-toolbox/shogun] Pull Request https://github.com/shogun-toolbox/shogun/pull/3903 opened by micmn	13:05
@wiking	micmn, thi sis the shiaatz? :)	13:06
@wiking	ok merging :D	13:06
@wiking	HeikoS, \o/	13:07
@wiking	no i meant	13:07
@wiking	HeikoS, \o/	13:07
@wiking	nono	13:07
@wiking	i meant!!!	13:07
@wiking	geektoni, \o/	13:07
@wiking	HeikoS, have u checked the file in TB?	13:07
@sukey	[https://github.com/shogun-toolbox/shogun] Pull Request https://github.com/shogun-toolbox/shogun/pull/3903 merged by vigsterkr	13:08
@sukey	[https://github.com/shogun-toolbox/shogun] New commit https://github.com/shogun-toolbox/shogun/commit/1ee49a8077961872b5125c7d68d4076e2f3add00 by vigsterkr	13:08
@wiking	boom boom	13:08
@wiking	micmn, rebase push on the lda pr?	13:08
-shogun-buildbot:#shogun- Build trusty - libshogun - viennacl #49 is complete: Failure [failed test (failure)] - http://buildbot.shogun-toolbox.org:8080/#builders/6/builds/49		13:10
@wiking	yeye this will be broken till we merge micmn stuff	13:10
@wiking	geektoni, it all popped up at one step?	13:12
geektoni	wiking: I think that the histogram shows the distribution of the weight's values	13:13
@wiking	geektoni, so now both histo and scalar is observed by ParameterObserverScalar	13:13
@wiking	?	13:13
geektoni	currently yes, but I was planning to refactor things a bit	13:14
@wiking	kk	13:14
@wiking	HeikoS, http://imgur.com/a/34en3	13:14
@wiking	lisitsyn, http://imgur.com/a/34en3	13:14
@HeikoS	prettty epic	13:15
@wiking	cool shiatz :)	13:16
@wiking	geektoni, way to go!	13:16
@wiking	geektoni, do you have twitter acc?	13:17
@wiking	ok found u	13:23
@wiking	:)	13:23
Trixis	wiking: ping	13:25
@wiking	Trixis, pong	13:26
@wiking	Trixis, so i've got some jnilib + jar for you	13:26
@wiking	you care to test it?	13:26
Trixis	wiking: awesome, sure!	13:26
Trixis	thx	13:26
@wiking	Trixis, mmm oh wait	13:26
@wiking	you are using shogun out of brew	13:26
@wiking	right?	13:26
Trixis	yeah	13:27
@wiking	okok	13:27
@wiking	lemme regenerate it	13:27
@wiking	:)	13:27
@wiking	HeikoS, pingu	13:30
@HeikoS	jo	13:30
@wiking	HeikoS, can u go through this: https://github.com/shogun-toolbox/shogun/pull/3901	13:31
@wiking	as soon i'm merging	13:31
@HeikoS	ok	13:31
@wiking	thnx	13:32
@wiking	micmn, no need for rebase push... i've just restarted travis ;)	13:33
micmn	mmm ok, but I updated the PR (clang-format and comments)	13:36
@sukey	[https://github.com/shogun-toolbox/shogun] Pull Request https://github.com/shogun-toolbox/shogun/pull/3901 synchronized by micmn	13:37
@wiking	hhahaha	13:39
@wiking	ok	13:39
@wiking	no worries	13:39
@wiking	lemme just kill	13:39
@wiking	the running tasks	13:39
@wiking	ok killed the tasks	13:41
@wiking	so your stuff gets priod	13:42
@wiking	micmn, you have a twitter acc?	13:44
@wiking	:D	13:44
@HeikoS	wiking: reviewed	13:45
@HeikoS	nice work micmn	13:45
@wiking	HeikoS, yep seen it! thnx heeps	13:45
@wiking	*heaps	13:45
micmn	thx!	13:46
@wiking	micmn, so twitter account ?:)	13:46
@wiking	yes or no? :)	13:46
micmn	mmm I think so	13:46
@wiking	kk i think you've just followed me	13:47
@wiking	ok	13:47
@wiking	where's the benchmarks gist? :D	13:47
@wiking	oh cool	13:48
micmn	https://gist.github.com/micmn/d4b51bf39676621afca563b55bdb9a98	13:48
@wiking	ye pyep	13:48
@wiking	boom	13:49
@wiking	:)	13:49
@wiking	HeikoS, ^	13:50
@HeikoS	nice one	13:51
@HeikoS	micmn: well done :)	13:51
@HeikoS	you think some stuff can be parallelized?	13:51
@HeikoS	micmn: btw which python scikit do you compare to	13:51
@HeikoS	because anaconda has multithreaded numpy	13:51
@HeikoS	should compare to that	13:51
@HeikoS	(even faster on single thread)	13:51
@HeikoS	I think some of the loops in the lda can be openmped?	13:52
@HeikoS	micmn: what would also be good that if a solver fails, then it falls back to the more stable one	13:52
@HeikoS	micmn: and then, BLOG about this, the world needs to know :D	13:52
@wiking	HeikoS, you and your anaconda	13:54
@wiking	:)	13:54
@wiking	there's a lot of people who dont use that :)	13:54
@wiking	really a lot :D	13:54
@HeikoS	wiking: well it IS like 15 times faster on my work computer	13:54
@HeikoS	I know	13:54
@wiking	yes	13:54
@HeikoS	but still	13:54
@wiking	but it is really not used	13:54
@wiking	by many	13:54
@wiking	this is again the 1%	13:54
@HeikoS	dont think	13:54
@HeikoS	used by almost everything at UCL	13:54
@wiking	ok ucl	13:55
@HeikoS	and the start ups here as well in london	13:55
@HeikoS	(the ones I know)	13:55
@HeikoS	we use it at swhere	13:55
@wiking	ok i have been around handful of startups	13:55
@HeikoS	client in norway uses it	13:55
@wiking	and other comapnies	13:55
@HeikoS	but I mean	13:55
@wiking	nobody uses it	13:55
@HeikoS	this is not even anaconda related	13:55
@HeikoS	more like MKL	13:55
@wiking	so how do we do this :)	13:55
@HeikoS	MKL will come to python at some point	13:55
@wiking	what is our stats here :)	13:55
@wiking	yeye but for mkl you need to fix cmake of shogun	13:55
@wiking	for eigen part	13:55
@HeikoS	yeah	13:55
@wiking	that's another story	13:56
@wiking	it's not complicated	13:56
@HeikoS	one of the gsocs has interest in that?	13:56
@wiking	it's just shitty	13:56
@wiking	i linked once the document	13:56
@HeikoS	micmn: ? that would speed up all of the eigen stuff in multicpore	13:56
@wiking	the env vars needs to be set	13:56
@wiking	on compile time	13:56
@wiking	HeikoS, you could actually ask	13:56
@HeikoS	so yeah I agree lets forget anaconda, but let's compare to MKL sklearn, and use MKL on our own	13:56
@HeikoS	micmn: nevermind, we do that later	13:58
@HeikoS	good code is more important	13:58
@HeikoS	wiking, micmn btw do we have like a list of shogun comparisons to sklearn?	13:58
@HeikoS	that is up to date etc?	13:58
@wiking	HeikoS, heheh for that we need actually rcurtin :)	13:58
micmn	HeikoS, wiking: eating, brb	13:59
@HeikoS	kk	13:59
@wiking	micmn, bonapetito	13:59
micmn	grazie :)	14:00
@wiking	Trixis, ok finally	14:04
@wiking	got it	14:04
@wiking	Trixis, lemme upload it to somewhere so you have it	14:04
Trixis	kk	14:04
-!- HeikoS [~heiko@host-92-0-169-11.as43234.net] has quit [Quit: Leaving.]		14:08
-!- HeikoS1 [~heiko@host-92-0-169-11.as43234.net] has joined #shogun		14:08
-shogun-buildbot:#shogun- Build deb3 - interfaces #19 is complete: Failure [failed test python (failure) test ruby (failure) test java (failure) test csharp (failure) test octave (failure) test R (failure)] - http://buildbot.shogun-toolbox.org:8080/#builders/37/builds/19		14:09
@wiking	jeje its ok	14:09
@wiking	Trixis, http://maeth.com/shogun-6.0.0/	14:09
@wiking	Trixis, and you'll need of course libshogun.18.0.dylib	14:10
@wiking	but that you'll have with brew	14:10
Trixis	yeah ill try replacing the corresponding brew files with these two and running the test code	14:11
@wiking	k	14:11
@wiking	can you paste me the test command	14:11
@wiking	you are trying	14:11
@wiking	plz	14:11
-shogun-buildbot:#shogun- Build deb1 - libshogun #50 is complete: Failure [failed test (failure)] - http://buildbot.shogun-toolbox.org:8080/#builders/10/builds/50		14:12
-!- HeikoS1 [~heiko@host-92-0-169-11.as43234.net] has quit [Ping timeout: 248 seconds]		14:13
Trixis	wiking: and it failed again, same error. The test is just a unit test that tries to call modshogun.init_shogun();	14:20
@wiking	oh ok	14:21
@wiking	you do it with ctest?	14:21
@wiking	or how do you exectute it	14:21
@wiking	*execute it	14:21
Trixis	wiking: this is in java, junit	14:22
@wiking	oooh	14:22
@wiking	you are running it with mvn?	14:22
Trixis	i tried compiling and running the equivalent snippet in C and that works	14:22
Trixis	wiking: ye	14:22
@wiking	eyye	14:22
@wiking	okok	14:22
@wiking	can u share the pom.xml?	14:22
@wiking	or just the part	14:23
@wiking	where you define the jnilib path	14:23
@wiking	because i'm just wondering wtf	14:23
@wiking	/usr/bin/java "-Xmx1024m" "-cp" "/usr/local/share/java/jblas.jar:/Users/wiking/shogun/build/src/interfaces/java_modular/shogun.jar:/Users/wiking/shogun/build/examples/meta/java:/Users/wiking/shogun/build/examples/meta/java/binary_classifier" "-Djava.library.path=/Users/wiking/shogun/build/src/interfaces/java_modular" "lda"	14:23
@wiking	this is in my case	14:23
@wiking	the command	14:23
micmn	wiking: [class_list problems] what's the C in class names? Can I avoid the inheritance from SGObject in internal classes like LDASolver?	14:44
@wiking	micmn, mmm if you wanna expose those to modular interface	14:45
@wiking	no	14:45
@wiking	meaming you wanna have LDASolver in python	14:45
@wiking	or java	14:45
micmn	no	14:45
@wiking	or any other none c++ itnerface	14:45
@wiking	if you just wanna expose the actual	14:45
@wiking	enum	14:45
@wiking	but not the implementatioj	14:45
@wiking	then you can omit SGObject inheritance	14:45
@wiking	unless	14:45
@wiking	wait	14:45
@wiking	unless :)	14:46
@wiking	this is actually a solver that never gets serialized	14:46
@wiking	i mean this is never part of a model	14:46
@wiking	the actual solver right?	14:46
micmn	right	14:46
@wiking	just being called	14:46
@wiking	okok	14:46
@wiking	you can omit then the SGObject inheritance	14:46
micmn	and the 'C'? :P	14:46
@wiking	yeah i mean	14:47
@wiking	that's a thing we wanna drop in the future	14:47
@wiking	so feel free to	14:47
@wiking	lalala	14:47
@wiking	:)	14:47
@wiking	;dancer:	14:47
@wiking	;)	14:47
micmn	thx :)	14:47
-!- sukey [~nodebot@ks312251.kimsufi.com] has quit [Remote host closed the connection]		14:50
-!- sukey [~nodebot@ks312251.kimsufi.com] has joined #shogun		14:51
-!- mode/#shogun [+o sukey] by ChanServ		14:51
Trixis	wiking: "-Djava.library.path=/usr/local/Cellar/shogun/6.0.0/lib/cli/shogun" is the only argument being passed to the VM (however shogun is also set up in intellij classpath so it should already be covered that way)	14:51
Trixis	wiking: thats the folder where the .jnilib resides	14:51
@wiking	Trixis, and where's your libshogun.so.18?	14:51
Trixis	that'd be /usr/local/Cellar/shogun/6.0.0/lib/	14:52
Trixis	set via DYLD_LIBRARY_PATH="/usr/local/Cellar/shogun/6.0.0/lib:$DYLD_LIBRARY_PATH"	14:52
@wiking	mmmm	14:55
@wiking	motherucker	14:55
Trixis	wiking: any idea what library is it trying to load? can't tell from just "java.lang.UnsatisfiedLinkError: org.shogun.modshogunJNI.init_shogun__SWIG_4()V". I was thinking that i could try loading the library thats causing the issue in my own snippet via System.load	14:58
@wiking	mmm how do you load the modshogun?	14:59
-!- slayerjain [uid117577@gateway/web/irccloud.com/x-rrracgzoqujmlrgb] has joined #shogun		15:04
Trixis	what exactly do you mean?	15:04
@wiking	static {	15:06
@wiking	System.loadLibrary("modshogun");	15:06
@wiking	}	15:06
@wiking	where do you load it	15:06
@wiking	in case of unit test	15:06
Trixis	i dont think im loading it manually	15:13
Trixis	i was thinking of trying to see if that'd work	15:13
@wiking	Trixis, you need to load that somewhere	15:14
@wiking	it wont load the jnilib otherwise in	15:14
@wiking	Trixis, have u check this doc http://www.swig.org/Doc3.0/Java.html#Java_dynamic_linking_problems	15:15
-!- iglesiasg [~iglesiasg@217.119.234.214] has quit [Quit: leaving]		15:28
@sukey	[https://github.com/shogun-toolbox/shogun] Pull Request https://github.com/shogun-toolbox/shogun/pull/3877 synchronized by geektoni	15:31
geektoni	wiking: ^	15:32
@wiking	geektoni, yep	15:34
@wiking	just trying to push the first official version of the tflogger	15:34
@wiking	i'm just as usual	15:34
@wiking	lost in the file/class naming helll	15:34
geektoni	wiking: cool!	15:35
geektoni	wiking: be creative! ;)	15:36
Trixis	wiking: hah, so loadLibrary doesnt work either w/ modshogun	15:39
Trixis	it's definitely in librarypath	15:39
Trixis	but it throws image not found for whatever reason	15:39
@wiking	mmm	15:40
@wiking	can you paste the error?	15:40
-!- iglesiasg [~iglesiasg@217.119.234.214] has joined #shogun		15:41
-!- mode/#shogun [+o iglesiasg] by ChanServ		15:41
Trixis	this is going to be spammy lol	15:41
@wiking	Trixis, pastebin.com	15:41
Trixis	wiking: https://pastebin.com/5k7t3BAg i have no idea what protobuf is	15:43
@wiking	ok so	15:43
@wiking	Trixis, have you saved	15:44
@wiking	your own jnilib?	15:44
Trixis	yes	15:44
@wiking	or you just overwrote?	15:44
Trixis	i saved it	15:44
@wiking	can you put that there ? /usr/local/Cellar/shogun/6.0.0/lib/cli/shogun/	15:44
@wiking	and try with that	15:44
@wiking	but you can have protobuf with brew install protobuf	15:44
Trixis	wiking: awesome, it runs now!	15:46
@wiking	Trixis, :D	15:46
Trixis	wiking: i assumed the shogun.jar would make sure modshogun was loaded, lol	15:46
Trixis	wiking: thanks a lot	15:46
@wiking	Trixis, unfortunately not yet	15:48
@wiking	nw	15:49
@sukey	[https://github.com/shogun-toolbox/shogun] Pull Request https://github.com/shogun-toolbox/shogun/pull/3901 synchronized by micmn	15:51
Trixis	wiking: my future questions will hopefully be more interesting, probably going to be related getting MKL SVMs to work, lol	15:54
@wiking	:)	15:54
@sukey	[https://github.com/shogun-toolbox/shogun] Pull Request https://github.com/shogun-toolbox/shogun/pull/3877 synchronized by geektoni	15:55
@sukey	[https://github.com/shogun-toolbox/shogun] Pull Request https://github.com/shogun-toolbox/shogun/pull/3901 merged by vigsterkr	16:38
@sukey	[https://github.com/shogun-toolbox/shogun] vigsterkr pushed 3 commits:	16:38
@sukey	https://github.com/shogun-toolbox/shogun/commit/fdfef6577a033a20b9feb0b30a6c0cb3af4eb58e	16:38
@sukey	https://github.com/shogun-toolbox/shogun/commit/e29ddfb1b7bae775adc26b89b73ec8c1e387097e	16:38
@sukey	https://github.com/shogun-toolbox/shogun/commit/00ec38a7ba141207e1c5ba42a0c3dcfa47ad4434	16:38
-shogun-buildbot:#shogun- Build trusty - libshogun - viennacl #50 is complete: Success [build successful] - http://buildbot.shogun-toolbox.org:8080/#builders/6/builds/50		16:44
-shogun-buildbot:#shogun- Build deb1 - libshogun #51 is complete: Success [build successful] - http://buildbot.shogun-toolbox.org:8080/#builders/10/builds/51		16:49
-!- iglesiasg [~iglesiasg@217.119.234.214] has quit [Quit: leaving]		17:07
-!- travis-ci [~travis-ci@ec2-54-158-166-19.compute-1.amazonaws.com] has joined #shogun		17:27
travis-ci	it's Michele Mazzoni's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: https://travis-ci.org/shogun-toolbox/shogun/builds/250378672	17:27
-!- travis-ci [~travis-ci@ec2-54-158-166-19.compute-1.amazonaws.com] has left #shogun []		17:27
-shogun-buildbot:#shogun- Build deb3 - interfaces #20 is complete: Success [build successful] - http://buildbot.shogun-toolbox.org:8080/#builders/37/builds/20		17:43
-!- geektoni [~geektoni@93-34-128-38.ip49.fastwebnet.it] has quit [Quit: Leaving.]		18:10
-!- HeikoS [~heiko@82-132-225-165.dab.02.net] has joined #shogun		18:20
-!- mode/#shogun [+o HeikoS] by ChanServ		18:20
@sukey	[https://github.com/shogun-toolbox/shogun] New commit https://github.com/shogun-toolbox/shogun/commit/846d0dd1244e96f91a2c40aab06595ce9cb3223a by karlnapf	18:24
@HeikoS	shogun-buildbot: force build 'nightly_bsd_license'	18:26
@HeikoS	lisitsyn: jo	18:38
lisitsyn	HeikoS: hey	18:38
-!- HeikoS [~heiko@82-132-225-165.dab.02.net] has quit [Ping timeout: 276 seconds]		18:43
-shogun-buildbot:#shogun- Build nightly_bsd_license #12 is complete: Success [build successful] - http://buildbot.shogun-toolbox.org:8080/#builders/4/builds/12		18:57
-!- slayerjain [uid117577@gateway/web/irccloud.com/x-rrracgzoqujmlrgb] has quit [Quit: Connection closed for inactivity]		18:59
-!- HeikoS [~heiko@82-132-215-133.dab.02.net] has joined #shogun		19:28
-!- mode/#shogun [+o HeikoS] by ChanServ		19:28
-!- HeikoS [~heiko@82-132-215-133.dab.02.net] has quit [Ping timeout: 255 seconds]		19:33
-!- shogun-buildbot [~shogun-bu@7nn.de] has quit [Remote host closed the connection]		19:45
-!- geektoni [~geektoni@93-34-128-38.ip49.fastwebnet.it] has joined #shogun		19:49
-!- shogun-buildbot [~shogun-bu@7nn.de] has joined #shogun		19:51
-!- geektoni [~geektoni@93-34-128-38.ip49.fastwebnet.it] has quit [Quit: Leaving.]		19:58
@sukey	[https://github.com/shogun-toolbox/shogun] New commit https://github.com/shogun-toolbox/shogun/commit/ad69234697cf50a38bbdbe1afa2de226963aafb3 by vigsterkr	20:02
@sukey	[https://github.com/shogun-toolbox/shogun] Issue https://github.com/shogun-toolbox/shogun/issues/3466 assigned to: vigsterkr by vigsterkr	20:08
-!- travis-ci [~travis-ci@ec2-54-158-166-19.compute-1.amazonaws.com] has joined #shogun		20:19
travis-ci	it's Heiko Strathmann's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: https://travis-ci.org/shogun-toolbox/shogun/builds/250456334	20:19
-!- travis-ci [~travis-ci@ec2-54-158-166-19.compute-1.amazonaws.com] has left #shogun []		20:19
@sukey	[https://github.com/shogun-toolbox/shogun] New commit https://github.com/shogun-toolbox/shogun/commit/ce9137e4c86513c8e2dc6ab187cbb23f82b1b06b by lisitsyn	21:23
@sukey	[https://github.com/shogun-toolbox/shogun] Issue https://github.com/shogun-toolbox/shogun/issues/3904 vigsterkr added label: "SWIG"	21:39
@sukey	[https://github.com/shogun-toolbox/shogun] Issue https://github.com/shogun-toolbox/shogun/issues/3904 assigned to: karlnapf by vigsterkr	21:39
@sukey	[https://github.com/shogun-toolbox/shogun] Issue https://github.com/shogun-toolbox/shogun/issues/3904 opened by vigsterkr	21:39
-!- mikeling [uid89706@gateway/web/irccloud.com/x-qlobivsojkbftbxq] has quit [Quit: Connection closed for inactivity]		22:28
-!- sukey [~nodebot@ks312251.kimsufi.com] has quit [Remote host closed the connection]		23:30
-!- sukey [~nodebot@ks312251.kimsufi.com] has joined #shogun		23:31
-!- mode/#shogun [+o sukey] by ChanServ		23:31
@sukey	[https://github.com/shogun-toolbox/shogun] Pull Request https://github.com/shogun-toolbox/shogun/pull/3905 opened by lisitsyn	23:37
-!- HeikoS [~heiko@host-80-43-248-60.as13285.net] has joined #shogun		23:40
-!- mode/#shogun [+o HeikoS] by ChanServ		23:40
-!- HeikoS [~heiko@host-80-43-248-60.as13285.net] has quit [Ping timeout: 240 seconds]		23:44
-!- HeikoS [~heiko@82-132-223-153.dab.02.net] has joined #shogun		23:45
-!- mode/#shogun [+o HeikoS] by ChanServ		23:45
@sukey	[https://github.com/shogun-toolbox/shogun] New commit https://github.com/shogun-toolbox/shogun/commit/728a9f57a65d462c42be0e487dd26a26545db9de by karlnapf	23:45
@sukey	[https://github.com/shogun-toolbox/shogun] Issue https://github.com/shogun-toolbox/shogun/issues/3904 closed by karlnapf	23:46
lisitsyn	HeikoS: hey	23:48
@HeikoS	lisitsyn: hi	23:48
lisitsyn	you wanted to tell me something?	23:48
@HeikoS	lisitsyn: yes	23:48
@HeikoS	license headers	23:48
@HeikoS	lisitsyn: need to automatically remove all GPL headers, replace with new one	23:49
lisitsyn	ahh	23:49
lisitsyn	that thing	23:49
@HeikoS	you had something there	23:49
lisitsyn	yes	23:49
@HeikoS	lisitsyn: could you	23:49
@HeikoS	a) double check whether really all gpl code has been moved	23:49
@HeikoS	b) change header of all files to BSD?	23:49
lisitsyn	HeikoS: yeah	23:50
lisitsyn	let me spend some time on that mañana	23:50
@HeikoS	cool	23:51
@HeikoS	lisitsyn: and then we can move code to other repo	23:51
@HeikoS	make cmake clone it	23:51
@HeikoS	and bam	23:51
@HeikoS	we are bsd	23:51
lisitsyn	super	23:51
lisitsyn	HeikoS: what about create? you have better name? ;)	23:51
@HeikoS	lisitsyn: was just thinking	23:51
@HeikoS	what does it do	23:52
@HeikoS	it creates an empty instance	23:52
lisitsyn	it spawns, creates	23:52
lisitsyn	instantiates :D	23:52
lisitsyn	whatever	23:52
@HeikoS	instantiate is good	23:52
@HeikoS	didnt like the sg?	23:52
@HeikoS	instantiate_sgoject	23:52
@HeikoS	long	23:52
lisitsyn	why?	23:52
@HeikoS	it is a question	23:52
@HeikoS	didnt like the sg?	23:52
lisitsyn	sgobject is in the return type	23:53
@HeikoS	I mean I dont care	23:53
lisitsyn	so we need a verb	23:53
@HeikoS	yeah true	23:53
lisitsyn	just a verb is fine I believe	23:53
lisitsyn	ok lets go with create this time	23:53
@HeikoS	instantiate is maybe a bit better, but I think create is ok	23:53
@HeikoS	its not like this is exposed to outside anyways	23:53
@HeikoS	but serializable is horrible :D	23:53
@HeikoS	Nice!	23:53
@HeikoS	you keep on cleaning	23:53
lisitsyn	HeikoS: actually gradually this thing becomes our interface	23:53
@HeikoS	is this like some long list or what? :D	23:53
lisitsyn	no I just do random stuff	23:54
lisitsyn	or?	23:54
@HeikoS	haha	23:54
lisitsyn	long list?	23:54
lisitsyn	which long list?	23:54
@HeikoS	the license thing would be ace	23:54
@HeikoS	but need to check first	23:54
@HeikoS	so we dont just re-declare gpl code as bsd	23:54
@HeikoS	lisitsyn: I'll go offline now	23:55
@HeikoS	have a good night!	23:55
lisitsyn	see you	23:55
lisitsyn	nite	23:55
-!- yamz [~myamada@2620:13:0:10:bdae:b37e:9286:1eb6] has joined #shogun		23:56
-!- HeikoS [~heiko@82-132-223-153.dab.02.net] has quit [Remote host closed the connection]		23:58
-!- yamz [~myamada@2620:13:0:10:bdae:b37e:9286:1eb6] has quit [Client Quit]		23:58
-!- yamz [~myamada@2620:13:0:10:bdae:b37e:9286:1eb6] has joined #shogun		23:59
--- Log closed Thu Jul 06 00:00:43 2017

Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!