Open in new window / Try shogun cloud
--- Log opened Mon Jun 27 00:00:57 2016
-!- arianepaola [~ariane@unaffiliated/arianepaola] has quit [Read error: Connection reset by peer]03:52
-!- arianepaola [~ariane@unaffiliated/arianepaola] has joined #shogun03:56
shogun-buildbotbuild #29 of memleak - valgrind is complete: Failure [failed memory check generate html]  Build details are at  blamelist: Heiko Strathmann <>, OXPHOS <>05:38
shogun-buildbotbuild #1030 of nightly_none is complete: Failure [failed test]  Build details are at  blamelist: Heiko Strathmann <>, OXPHOS <>05:43
shogun-buildbotbuild #1160 of nightly_default is complete: Failure [failed test test_1 notebooks]  Build details are at  blamelist: Heiko Strathmann <>, OXPHOS <>06:27
arianepaolaping wiking08:16
BZDMGbazdmeg morning08:33
@wikingmorning :)08:33
-!- sanuj [~sanuj@] has joined #shogun09:40
sanujBZDMG, there?09:40
sanujI solved the bug09:41
sanujvery silly bug09:41
arianepaolaping wiking10:26
BZDMGso what's the bug10:55
sanujone word10:55
sanujshould be void set(...)10:56
sanuji actually did all the changes step by step on a fresh branch to spot this10:57
-!- OXPHOS [c68f0c0c@gateway/web/freenode/ip.] has joined #shogun11:10
-!- arianepaola [~ariane@unaffiliated/arianepaola] has quit [Read error: Connection reset by peer]11:14
-!- HeikoS [] has joined #shogun11:26
-!- mode/#shogun [+o HeikoS] by ChanServ11:26
@wikingHeikoS: yo11:27
@wikingOXPHOS: china!11:27
OXPHOSwiking: hello!11:27
@wikingOXPHOS sanuj Saurabh7__11:28
@wikingwhere is the weekly email? :)11:28
@wikingOXPHOS sanuj Saurabh7__ PING PING PING PING PING11:29
@wikingplease let's try to have this in your calendars11:29
@wikingso i dont have to do this every week11:29
OXPHOSwiking: working on it. trying to locate the revised schedule i wrote yesterday..11:29
@wikingit's 5:28pm11:29
@wikingi know where u r11:30
@HeikoSin fact, as I said in the last email, pls also send a 1-2 sentence short status every day you work on shogun11:30
sanujokay, sending11:30
sanuji was thinking of reporting what i had done today11:30
sanujsince the day is not over, i didn't send11:30
@HeikoSthe daily thing can me minor, totally fine11:30
@HeikoSjust for us to keep track11:30
sanujsending to HeikoS BZDMG wiking and lambday11:31
@wikingmailing list11:31
sanujshogun list11:32
@wikingis a perfect place11:32
@HeikoSsanuj: so what sup with da bug?11:32
sanujHeikoS, solved it11:32
sanujwas silly11:32
@HeikoSsanuj: great!!11:32
@HeikoSsanuj: what was it?11:32
sanujHeikoS, i actually did all the changes step by step on a fresh branch to spot this as you had suggested ;)11:32
sanujshould have been void set(...)11:33
sanujforgot to do this ^11:33
OXPHOSwiking: haha you caught me11:33
@HeikoSsanuj: this is the private method to add a nonexisting tag?11:34
@HeikoSsanuj: can you rename this to something better then _set?11:34
sanujno this is in Self struct in SGObject.cpp11:34
sanujHeikoS, i have renamed _set to add() to set a parameter for first time from init_param()11:35
@HeikoSsanuj: cool11:35
@HeikoSso then, you can re-add the things you had developed11:35
@HeikoSor did you already?11:35
@HeikoSsanuj: and then we can finally merge this pr11:35
sanujHeikoS, i'm just adding some unit-tests11:35
sanujon MockObject11:35
@HeikoSsanuj: great, pls ping me once done11:35
@HeikoSI am online for 2 more hours or so11:36
sanujHeikoS, that's good11:36
@HeikoSOXPHOS: so how are things going?11:36
sanujback in UK?11:36
sanujno it seems11:36
@HeikoSsanuj: no thats my workstation in the office11:36
@HeikoSI am still hiding11:36
sanujyeah....that's how in inferred that you are not in london :P11:36
OXPHOSHeikoS: good! revising based on your comments11:37
@HeikoSOXPHOS: ok, as said, I am here for a while now11:37
@HeikoSso we can discuss also, if needed11:37
@HeikoSwiking: when do you finally tell us how to get rid of the templating macros?11:37
OXPHOSHeikoS: cool! I have a question -11:39
OXPHOStemplate<class T>   void SGVector<T>::copy_data(const SGReferencedData &orig)   { gpu_vector=std::shared_ptr<GPUMemoryBase<T>>(((SGVector*)(&orig))->gpu_vector);11:39
OXPHOSyou said: this crashes if I pass SGMatrix, no? If so, this was a bug. Could need a fix there11:39
@HeikoSyeah since you static cast11:39
@HeikoSand if I passed something else than SGVector what happens?11:40
@HeikoSOXPHOS: ever saw this:
OXPHOSlike some other derived of SGRef?11:40
@HeikoSOXPHOS: yes, like SGMatrix11:41
@HeikoSno type error11:41
@HeikoSbut static cast will fail11:41
OXPHOSHeikoS: yeah i see. But not sure how to avoid that..?11:43
@HeikoSOXPHOS: can you give me the link again?11:43
OXPHOSHeikoS: is_same also requires a list of all types right? why is it different from the template<T> stuff?11:43
@HeikoSOXPHOS: not sure, just was looking at it, never really checked it before11:44
@HeikoSOXPHOS: so copy_data is not a method of SGReferencedData, right? Just of SGVector?11:45
OXPHOSHeikoS: it's pure virtual in SGRef11:46
@HeikoSOXPHOS: I see11:46
@HeikoSok then11:46
OXPHOSBut if we also have a gpu_data in SGMatrix maybe it is okay..?11:46
@HeikoSI kind of dont wanna put a dynamic cast11:46
@HeikoSso just leave it for now11:46
@HeikoSIll ask BZDMG11:46
OXPHOSsure HKS!11:47
@HeikoSBZDMG: copy_data method11:49
@HeikoSpurely virtual in SGReferencedData11:49
-!- lambday [8028b10a@gateway/web/freenode/ip.] has joined #shogun11:49
-!- mode/#shogun [+o lambday] by ChanServ11:49
@HeikoSimplemented in SGVector11:49
@HeikoSBZDMG: doesnt check type11:49
@HeikoSbut static casts to SGVector11:49
@lambdayerm.. what are we talking about?11:50
@HeikoSlambday: minor stuff11:50
@HeikoSlambday: whats needed to merge OXPHOS PR?11:50
@lambdayHeikoS: let me check11:50
@lambdaywell the build is broken.11:51
@lambdaysecondly, this still is for develop branch11:51
OXPHOSDo I suppose to split them into .h and .cpp?11:51
@lambdayOXPHOS: why didn't you just refactor things in the linalg_refactor branch? there you've already added the global instance11:51
@lambdayOXPHOS: why do we need to keep adding the same thing in multiple branches? :)11:52
BZDMGHeikoS: show me the code I don't get it :)11:52
OXPHOSlambday: dunno..because some stuff are already merged?11:52
OXPHOSlambday: I can pull it against that branch for this update11:53
@lambdayOXPHOS: merged to the feature branch.. doesn't mean we can't change things later :)11:53
OXPHOSlambday: haha sure. also why it always fails some tests?11:53
@lambdayOXPHOS: you'll have to check that11:54
@lambdayOXPHOS: but please make the PR against feature/linalg_refactor branch..11:54
@lambdaywhere you already added those stuffs.. just refactor it11:54
OXPHOSlambday: sure11:54
OXPHOSBZDMG: i think you're @-ed on Github for that line of codes11:55
@lambdayOXPHOS: also, I don't think you have turned off *all* the SGVector operations when data is on gpu.. we have to make sure that they throw error message instead of just crashing11:55
Saurabh7__HeikoS: yo11:56
OXPHOSlambday: not all..there're some static ones11:56
@lambdayOXPHOS: but in case you did that already, just ignore my comments :D11:56
OXPHOSlambday: those static methods..I think they should be moved out finally..and they takes T* not SGVectors actually so I think they're harmless for now11:58
@HeikoSSaurabh7__: jojo11:58
@lambdayOXPHOS: what are those methods?11:58
OXPHOSlambday: sum..add..norm..11:59
@lambdayOXPHOS: yeah they should be moved out! put to linalg11:59
@lambdayOXPHOS: but let's not do it all in just one patch11:59
Saurabh7__HeikoS: I think I have updated most comments in the PR,11:59
@HeikoSSaurabh7__: you feel it is merge ready?12:00
@HeikoSwill check soon12:00
OXPHOSlambday: sure12:00
@lambdayOXPHOS: it's easier to miss things - we gotta do it in some later patch12:00
Saurabh7__HeikoS: yes12:00
@lambdayBZDMG: who are your vowels?12:00
Saurabh7__mm lemme check memleak12:00
@HeikoSlambday: jo12:03
@HeikoSlambday: can you give me a little tip here12:03
@lambdayHeikoS: sure.. what is it about?12:04
@HeikoSlambday:  see pm12:04
@HeikoSlambday: can you see?12:05
@HeikoSSaurabh7__: got some timings on which changes helped?12:06
Saurabh7__HeikoS: yes will have to put them in12:08
@HeikoSwhat about this?12:19
Saurabh7__HeikoS: ah yes, a column is chosen there to go through12:20
Saurabh7__HeikoS: so do we need num_subset of columns12:21
Saurabh7__* we need12:21
Saurabh7__HeikoS: so you recommend sorting the first num_subset elements ?12:22
@HeikoSeverything that determines the order in which you access memory12:22
@HeikoSshould be sorted12:22
@HeikoSSaurabh7__: all my comments were minor12:22
@HeikoSIll give me OK to merge12:22
Saurabh7__HeikoS: ok thanks, will get those values out and do memcheck here12:23
@HeikoSSaurabh7__: cool12:27
@HeikoSSaurabh7__: once thats done, then we do get rid of the float64_t?12:27
Saurabh7__HeikoS: yes could be12:27
-!- shogun-notifier- [] has joined #shogun12:28
shogun-notifier-shogun-data: Saurabh7 :master * e27433b / testsuite/meta/ (2 files):
shogun-notifier-shogun-data: update rf data12:28
shogun-notifier-shogun-data: Heiko Strathmann :master * b9b2d90 / testsuite/meta/ (2 files):
shogun-notifier-shogun-data: Merge pull request #111 from Saurabh7/rfnewdata12:28
shogun-notifier-shogun-data: update rf data12:28
Saurabh7__right now its expected everything is already numerals12:28
-!- sanuj [~sanuj@] has quit [Ping timeout: 244 seconds]12:32
@HeikoSOXPHOS: jo12:35
@HeikoScould you clean up the doc?12:35
@HeikoSlots of deprecated stuff in there12:35
@HeikoSOXPHOS: maybe just create a new doc with the revised timeline12:35
@HeikoSnot sure where to look12:35
OXPHOSHeikoS: which one?12:36
@HeikoSthe doc you sent in the mail to list12:37
OXPHOSHeikoS: oh i see. so I'll remove the outdated ones12:37
@HeikoSOXPHOS, Saurabh7__, gotta leave soon12:50
@HeikoSmight be back later12:50
@HeikoSbut anything you want to discuss?12:50
OXPHOSHeikoS: not yet. thx!12:51
@HeikoScool see you later then12:52
-!- sanuj [~sanuj@] has joined #shogun12:53
Saurabh7__HeikoS: ok12:55
Saurabh7__HeikoS: next I will try multithreaded thing and removing float64_t ?12:55
@HeikoSSaurabh7__: start with float6412:56
@HeikoSthere might be some mini issues we need to solve with this12:56
@HeikoSor multithread12:56
@HeikoSyou can choose :)12:56
-!- HeikoS [] has quit [Quit: Leaving.]12:57
sanujBZDMG, there?13:11
BZDMGsanuj: a bit13:11
sanujBZDMG, need to discuss one thing13:11
sanujBZDMG, now params are set in init_params() for each class13:11
sanujthey are declared but not initialized with a value13:11
sanujBZDMG, if someone does get() then it will result in garbage value13:12
sanujwithout setting a value to it13:12
sanujBZDMG, ?13:33
BZDMGsanuj: get should fail13:33
BZDMGif there is no such thing13:34
BZDMGno garbage13:34
sanujBZDMG, currently we have this13:34
sanuj    void init_params()13:34
sanuj    {13:34
sanuj        m_vector = SGVector<float64_t>();13:34
sanuj        add("vector", m_vector);13:34
sanuj        add("int", m_int);13:34
sanuj        add("float", m_float);13:34
sanuj    }13:34
sanuj    SGVector<float64_t> m_vector;13:34
sanuj    int32_t m_int;13:34
sanuj    float64_t m_float;13:34
sanujBZDMG, how to make get() fail?13:34
sanujBZDMG, and we need to consider member variables with default values also....13:35
BZDMGBZDMG: get = if has(..) return else fail13:38
BZDMGsanuj: ^13:38
BZDMGnot really get the problem13:38
sanujBZDMG, there are two functions now in SGObject.h13:38
sanujset() and add()13:38
sanujadd is protected and set is public13:39
sanujadd("vector", m_vector); will do => map[basetag] = any13:39
sanujbut m_vector is empty13:39
sanujsame is the case for m_int, m_float13:39
sanujso m_int and m_float have random values since they are not initialized13:40
sanujBZDMG, has() will also return true since they exist in the map13:40
sanujBZDMG, got it?13:42
BZDMGsanuj: probably yes13:57
BZDMGbut it is not garbage13:57
BZDMGits default13:57
BZDMGI think we should explicitly provide what value to put13:57
sanujBZDMG, most members dont have default value13:57
BZDMGthey all have13:57
BZDMGit is either zero or something like that13:58
sanujBZDMG, no worries then13:58
-!- sanuj [~sanuj@] has quit [Ping timeout: 276 seconds]14:12
-!- sanuj [~sanuj@] has joined #shogun14:38
sanujBZDMG, one more question14:40
-!- OXPHOS [c68f0c0c@gateway/web/freenode/ip.] has quit [Quit: Page closed]15:27
-!- shogun-notifier- [] has quit [Quit: transmission timeout]15:28
-!- sanuj [~sanuj@] has quit [Ping timeout: 272 seconds]16:11
-!- sonne|osx [] has joined #shogun17:02
-!- lambday [8028b10a@gateway/web/freenode/ip.] has quit [Quit: Page closed]17:25
-!- sanuj [~sanuj@] has joined #shogun17:26
-!- sonne|osx [] has quit [Quit: sonne|osx]17:30
-!- sonne|osx [] has joined #shogun17:32
-!- sonne|osx [] has quit [Quit: sonne|osx]17:41
-!- lambday [8028b10a@gateway/web/freenode/ip.] has joined #shogun17:43
-!- mode/#shogun [+o lambday] by ChanServ17:43
-!- sonne|osx [] has joined #shogun17:56
-!- sonne|osx [] has quit [Client Quit]17:57
-!- lambday [8028b10a@gateway/web/freenode/ip.] has quit [Ping timeout: 250 seconds]18:46
-!- sonne|osx [] has joined #shogun19:06
-!- sonne|osx [] has quit [Client Quit]19:07
-!- sonne|osx [] has joined #shogun19:08
-!- sonne|osx [] has quit [Quit: sonne|osx]19:21
-!- sanuj [~sanuj@] has quit [Remote host closed the connection]20:22
-!- sonne|osx [] has joined #shogun21:12
-!- sonne|osx [] has quit [Quit: sonne|osx]21:18
-!- sonne|osx [] has joined #shogun21:28
-!- sonne|osx [] has quit [Quit: sonne|osx]21:40
-!- sonne|osx [] has joined #shogun22:11
-!- sonne|osx [] has quit [Quit: sonne|osx]22:25
-!- sonne|osx [] has joined #shogun22:27
-!- HeikoS [~heiko@] has joined #shogun22:29
-!- mode/#shogun [+o HeikoS] by ChanServ22:29
-!- sonne|osx [] has quit [Client Quit]22:29
-!- HeikoS [~heiko@] has quit [Quit: Leaving.]22:46
-!- shogun-notifier- [] has joined #shogun22:46
shogun-notifier-shogun: Heiko Strathmann :feature/kernel_exp_family * b027178 / / (3 files):
shogun-notifier-shogun: add nystrom version of xi_norm22:46
-!- travis-ci [] has joined #shogun23:22
travis-ciit's Heiko Strathmann's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun:
-!- travis-ci [] has left #shogun []23:22
--- Log closed Tue Jun 28 00:00:58 2016