Open in new window / Try shogun cloud
--- Log opened Sat Jun 30 00:00:17 2012
-!- blackburn [~blackburn@81.28.187.0] has joined #shogun00:02
-!- blackburn [~blackburn@81.28.187.0] has quit [Quit: Leaving.]00:06
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: leaving]00:14
-!- blackburn [~blackburn@81.28.187.0] has joined #shogun00:27
-!- blackburn [~blackburn@81.28.187.0] has quit [Ping timeout: 240 seconds]02:12
alexlovesdatabye all03:20
-!- alexlovesdata [55b21452@gateway/web/freenode/ip.85.178.20.82] has quit [Quit: Page closed]03:20
-!- gsomix [~gsomix@109.169.240.118] has joined #shogun07:52
-!- gsomix [~gsomix@109.169.240.118] has quit [Ping timeout: 250 seconds]08:42
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun10:49
-!- heiko [~heiko@host86-180-216-139.range86-180.btcentralplus.com] has joined #shogun11:00
-!- heiko [~heiko@host86-180-216-139.range86-180.btcentralplus.com] has quit [Ping timeout: 264 seconds]12:12
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Ping timeout: 265 seconds]12:36
-!- blackburn [~blackburn@81.28.187.0] has joined #shogun13:53
blackburnsonney2k: around?14:14
@sonney2kwassup14:41
@sonney2kblackburn, if you can please fix the buildbot errors - this is really highest prio now...14:46
blackburnsonney2k: sparse w/ python?14:47
blackburnI got stucked there a little yesterday14:47
blackburnsonney2k: can I ask you some debian things?14:47
blackburnok leaving now for a while - will ask later :)14:53
@sonney2kblackburn, where did you get stuck?15:25
@sonney2kwhat is the problem?15:25
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun15:57
--- Log closed Sat Jun 30 16:34:24 2012
--- Log opened Sat Jun 30 16:34:35 2012
-!- shogun-toolbox [~shogun@7nn.de] has joined #shogun16:34
-!- Irssi: #shogun: Total of 8 nicks [1 ops, 0 halfops, 0 voices, 7 normal]16:34
-!- Irssi: Join to #shogun was synced in 6 secs16:34
--- Log closed Sat Jun 30 16:40:42 2012
--- Log opened Sat Jun 30 16:46:53 2012
-!- shogun-toolbox [~shogun@7nn.de] has joined #shogun16:46
-!- Irssi: #shogun: Total of 9 nicks [1 ops, 0 halfops, 0 voices, 8 normal]16:46
-!- Irssi: Join to #shogun was synced in 6 secs16:46
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Read error: Operation timed out]17:18
-!- puffin444 [62e3926e@gateway/web/freenode/ip.98.227.146.110] has joined #shogun18:22
blackburnsonney2k: with detecting error18:39
@sonney2kblackburn, ?18:46
blackburnsonney2k: yesterday I got stucked with detecting error18:46
@sonney2kdoesn't valgrind help?18:46
@sonney2kI mean it must be due to the refcounting business18:46
blackburnyes it helps but I didn't localized the error18:47
blackburnsonney2k: how can I get bootable usb for installing debian?18:47
@sonney2kdon't know out of my head18:48
blackburnthey provide some crazy ways of installing it in debian doc18:48
blackburnis that really so complex?18:48
blackburnok probably live version can be copied via dd18:54
puffin444Hey blackburn can I ask you a question.18:55
blackburnsure18:55
puffin444I'm trying to think of ways to reduce computational redundancy in my GP code, so I store kernel matrices and other calcuated matrices until parameters are changed.18:56
puffin444This way they are not recalculated. A problem with this though is that when I use Heiko's model selection scheme to set parameters, the CMachine has no idea that parameters have been changed.18:57
puffin444Do you what caching/matrix storage schemes have been used in the past to overcome these things? Heiko may have more information, but he is not here at the moment.18:58
blackburnuhh18:58
puffin444Don't worry about it :)18:59
blackburnlet me try to get into it18:59
blackburnso you work with CMachines18:59
puffin444Yes.18:59
blackburnand you want to know when parameters are changed?19:00
puffin444Exactly.19:00
blackburnI don't see how to get it done currently19:01
blackburnpuffin444: that parameters framework actually don't know about changing parameters too19:02
@sonney2kpuffin444, what you could do is extend the framework and e.g. compute the md5sum of all parameters19:03
@sonney2kif that sum changes19:03
puffin444Would it be bad to add some abstract update() function to CMachine? Or is that just a lazy solution?19:03
@sonney2k-> parameters changed19:03
blackburnthat would take some time for you puffin44419:04
blackburnbut actually it is worth I think19:04
blackburnvoid fire_parameter_changed(const char* name)19:05
puffin444Is there an md5sum algorithm already in shogun, or would I have to use an external library/implement it myself?19:06
@sonney2kpuffin444, yes there is19:06
blackburnwhy to compute md5?19:06
@sonney2klook at CHash19:06
@sonney2kblackburn, to see if parameters change19:07
blackburnsonney2k: well if set_* is called we can assume it is changed19:07
@sonney2kwhy?19:07
blackburnsonney2k: are you on debian sid now?19:08
puffin444Okay. So would it make sense to add an md5sum member in SGObject that stores the md5sum of its model selection parameter tree?19:08
@sonney2kblackburn, yes19:10
puffin444Or would simply a function that calculates an md5sum of a parameter tree suffice?19:10
@sonney2kpuffin444, heiko is the expert here19:10
@sonney2knot sure what is best19:10
puffin444okay.19:11
blackburnpuffin444: why do you need to work with parameters at all?19:16
blackburnjust curious19:16
puffin444I use the model selection framework to do hyperparameter learning. Whenever one of those parameter changes I need to recalcuate the matrices I need.19:17
blackburnwhy is it so general?19:17
puffin444The gradient search was written to be used with arbitrary machines.19:19
blackburnI see19:19
blackburnhah nice no wi-fi driver in debian live19:26
blackburnno display driver as well19:26
blackburnand I have no cable19:26
@sonney2kpuffin444, btw I would go for a m_parameter_checksum in SGObject for the time being19:36
puffin444Okay. I think I can write an MD5 sum function using CHash.19:36
blackburnwill that approach require you to add anything to setters?19:37
@sonney2kpuffin444, alternatively you can also use murmurhash2 (also in CHash) should be good enough I think19:37
blackburnhmm it failes here to link with nlopt for some reason19:50
blackburnI guess I compiled nlopt w/o shared19:51
blackburnsonney2k: hmm I just got I do what I did yesterday20:21
blackburnprobably my commit went to hell20:21
blackburnI was really sure I pushed it20:21
-!- heiko [~heiko@host86-181-9-226.range86-181.btcentralplus.com] has joined #shogun20:27
-!- heiko [~heiko@host86-181-9-226.range86-181.btcentralplus.com] has quit [Ping timeout: 264 seconds]20:34
@sonney2kwiking, any news on your PR?20:37
CIA-18shogun: Sergey Lisitsyn master * ra10eab1 / (3 files in 3 dirs): Fixed sparse matrix handling in static - http://git.io/356OIw20:39
blackburnsonney2k: found it in backups :D20:39
blackburnkind of mystic20:39
blackburnsonney2k: do you know where one should report if kernel does not detect whether AC adapter changes status?20:47
@sonney2kblackburn, reportbug <packagename>21:56
@sonney2kblackburn, or lkml21:56
blackburnno idea about package name :D21:56
blackburnI tried everything21:56
blackburnstill acpi does not fire that AC is plugged21:57
blackburnstate is being set correctly21:57
blackburnin /proc/acpi/battery21:57
blackburnthe only bug left with my new notebook (I spent some time forcing it to produce lesser heat)21:58
@sonney2kblackburn, did you try to compile your own 3.4 / 3.5-rc kernel to see if it is fixed there?22:19
blackburnsonney2k: I installed 3.4 from kernel-ppa - no help here22:20
@sonney2kblackburn, then send an email to lkml asking for help22:20
blackburnsonney2k: problem is I am unsure it is a kernel bug22:20
blackburnsonney2k: /proc/ stuff updates perfectly - but acpi doesn't catch it22:21
blackburnmay be I should try newer acpid?22:21
blackburnstill the same with newer acpid argh22:33
@sonney2kcat /proc/acpi/ac_adapter/ADP1/state22:35
@sonney2kstate:                   off-line22:35
@sonney2kcat /proc/acpi/ac_adapter/ADP1/state22:35
@sonney2kstate:                   on-line22:35
@sonney2kblackburn, ^ it is all in kernel stuff22:36
@sonney2kso kernel bug22:36
blackburnsonney2k: that thing is being updated22:36
blackburnsonney2k: run acpi_listen and plug in / plug out your AC22:36
blackburnyou should see some msgs here - I do not22:36
@sonney2kI don't have acpid installed22:37
blackburnhmm22:37
blackburnsonney2k: how does it detect whether you are running AC??22:37
@sonney2kwhy should it detect it?22:38
blackburnwell for gnome22:38
@sonney2ksome gnome stuff no idea22:38
blackburnand for laptop mode22:38
@sonney2kI think upowerd does that22:38
@sonney2kwhich laptop mode?22:39
blackburnsonney2k: laptop-mode-tools package22:39
@sonney2kuseless stuff...22:39
blackburnsonney2k: well it controls cpufreqd22:39
@sonney2kwhy that22:39
@sonney2kjust use in kernel on-demand22:39
blackburnsonney2k: hmm well it has its own configuration22:40
@sonney2ksnakeoil22:41
blackburnmay be it could be a good idea to use only cpufreqd22:41
@sonney2knot even that22:43
@sonney2kjust use ondemand in kernel22:43
@sonney2k*always*22:43
blackburnsonney2k: in kernel??22:43
blackburnhow do you handle that in kernel?22:44
@sonney2kecho ondemand > /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor22:44
@sonney2k(and all your other cpu's)22:44
blackburnsonney2k: how do you make it persistent?22:44
@sonney2kput it in some init script22:45
blackburnsonney2k: do you have anything in /etc/cpufreqd.conf?22:47
@sonney2kI don't have that installed23:12
@sonney2knite23:12
-!- gsomix [~gsomix@109.169.232.38] has joined #shogun23:13
-!- gsomix [~gsomix@109.169.232.38] has quit [Client Quit]23:15
-!- puffin444 [62e3926e@gateway/web/freenode/ip.98.227.146.110] has quit [Quit: Page closed]23:42
--- Log closed Sun Jul 01 00:00:17 2012