--- Log opened Fri Jul 06 00:00:17 2012 | ||
blackburn | sonney2k: I failed to notice fluctuations in confs | 00:00 |
---|---|---|
blackburn | aham! found one | 00:00 |
@sonney2k | well nr +1= 2979 fluctuates | 00:01 |
@sonney2k | that cannot happen... | 00:01 |
nietpiet | I tried to monitor the process with top and it seemed single threaded, but the kernel loading takes most of the time so it is hard to see, i can print some debug output if that helps? (blackburn taught me how :) ) | 00:02 |
nietpiet | I can make the C value a bit higher so it takes longer? Easier to monitor with top :) | 00:03 |
blackburn | nietpiet: one thread will be used is you set it | 00:03 |
nietpiet | oh.. and i just print the first 100 confidence scores.. i could print more? | 00:04 |
@sonney2k | nietpiet, I would prefer if you store them and compute numpy.max(abs(out1-out2)) | 00:05 |
nietpiet | ok | 00:05 |
@sonney2k | nietpiet, please upload your script to dpaste | 00:07 |
@sonney2k | I simply cannot imagine that it doesn't give exact same results | 00:07 |
CIA-18 | shogun: Soeren Sonnenburg master * r463293d / src/shogun/machine/KernelMachine.cpp : enably apply() w/o arguments to work with custom kernels - http://git.io/0hjzIA | 00:10 |
@sonney2k | blackburn, ^ I hope that fixes it - nietpiet can you try? | 00:10 |
* sonney2k argh | 00:10 | |
@sonney2k | it is getting too late | 00:10 |
@sonney2k | I need sleep | 00:10 |
@sonney2k | cu | 00:10 |
nietpiet | @sonney2k, I will upload it, but it is a bit messy now, as you can imagine.. http://pastebin.com/EJ7nrM5z | 00:11 |
nietpiet | numpy.max(abs(out1-out2)) for the consequative runs gives 0.0 ; 1.0 ; 1.0 ; 2.00000023656 | 00:12 |
shogun-buildbot | build #65 of deb1 - libshogun started, including [463293d2198c9a5d510bb95f39c7c555af2b4b4d] | 00:12 |
shogun-buildbot | build #65 of deb1 - libshogun is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb1%20-%20libshogun/builds/65 | 00:13 |
shogun-buildbot | build #75 of deb2 - static_interfaces started, including [463293d2198c9a5d510bb95f39c7c555af2b4b4d] | 00:13 |
shogun-buildbot | build #76 of deb3 - modular_interfaces started, including [463293d2198c9a5d510bb95f39c7c555af2b4b4d] | 00:13 |
nietpiet | blackburn, um.. should i get a new version from git? | 00:13 |
n4nd0 | sonney2k: good night | 00:13 |
blackburn | nietpiet: well it must fix exactly the same thing you fixed before | 00:13 |
nietpiet | @sonney2k, Thank you very much for your help! and good night. | 00:14 |
shogun-buildbot | build #75 of deb2 - static_interfaces is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb2%20-%20static_interfaces/builds/75 | 00:19 |
nietpiet | Blackburn: I'm also calling it quits. Thank you for your help all day! | 00:24 |
blackburn | you are welcome | 00:24 |
blackburn | let's try to continue tomorrow | 00:24 |
nietpiet | yes, i'll be online, but probably less long. Good night. | 00:25 |
blackburn | nite | 00:26 |
-!- zxtx [~zv@cpe-75-83-151-252.socal.res.rr.com] has quit [Ping timeout: 245 seconds] | 00:33 | |
-!- nietpiet [9232901a@gateway/web/freenode/ip.146.50.144.26] has quit [Quit: Page closed] | 00:34 | |
shogun-buildbot | build #76 of deb3 - modular_interfaces is complete: Failure [failed test python_modular] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/76 blamelist: Soeren Sonnenburg <sonne@debian.org> | 01:02 |
CIA-18 | shogun: Sergey Lisitsyn master * r991fbc7 / (3 files in 2 dirs): Added custom index block tree creation by adjacency matrix - http://git.io/9d-BPA | 01:44 |
shogun-buildbot | build #66 of deb1 - libshogun started, including [991fbc74a44f89c7ae2e0ae08bc1e6cf9341455a] | 01:47 |
shogun-buildbot | build #66 of deb1 - libshogun is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb1%20-%20libshogun/builds/66 | 01:48 |
shogun-buildbot | build #77 of deb3 - modular_interfaces started, including [991fbc74a44f89c7ae2e0ae08bc1e6cf9341455a] | 01:48 |
shogun-buildbot | build #76 of deb2 - static_interfaces started, including [991fbc74a44f89c7ae2e0ae08bc1e6cf9341455a] | 01:48 |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: leaving] | 01:49 | |
shogun-buildbot | build #76 of deb2 - static_interfaces is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb2%20-%20static_interfaces/builds/76 | 01:53 |
CIA-18 | shogun: Sergey Lisitsyn master * r312d90a / src/shogun/lib/IndexBlockTree.cpp : Updated IndexBlockTree constructor routine - http://git.io/E42gAA | 02:35 |
shogun-buildbot | build #77 of deb3 - modular_interfaces is complete: Failure [failed test python_modular] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/77 blamelist: Sergey Lisitsyn <lisitsyn.s.o@gmail.com> | 02:36 |
shogun-buildbot | build #67 of deb1 - libshogun started, including [312d90a08e82a5f1e18f457381fb633edab8b8a7] | 02:37 |
shogun-buildbot | build #67 of deb1 - libshogun is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb1%20-%20libshogun/builds/67 | 02:38 |
shogun-buildbot | build #77 of deb2 - static_interfaces started, including [312d90a08e82a5f1e18f457381fb633edab8b8a7] | 02:38 |
shogun-buildbot | build #78 of deb3 - modular_interfaces started, including [312d90a08e82a5f1e18f457381fb633edab8b8a7] | 02:38 |
-!- blackburn [~blackburn@81.28.187.0] has quit [Quit: Leaving.] | 02:39 | |
shogun-buildbot | build #77 of deb2 - static_interfaces is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb2%20-%20static_interfaces/builds/77 | 02:43 |
shogun-buildbot | build #14 of nightly_default started, including [] | 03:00 |
shogun-buildbot | build #78 of deb3 - modular_interfaces is complete: Failure [failed test python_modular] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/78 blamelist: Sergey Lisitsyn <lisitsyn.s.o@gmail.com> | 03:28 |
shogun-buildbot | build #9 of nightly_all started, including [] | 03:28 |
shogun-buildbot | build #14 of nightly_default is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/nightly_default/builds/14 | 03:38 |
shogun-buildbot | build #12 of nightly_none started, including [] | 03:38 |
shogun-buildbot | build #12 of nightly_none is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/nightly_none/builds/12 | 03:50 |
shogun-buildbot | build #9 of nightly_all is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/nightly_all/builds/9 | 03:52 |
-!- puffin444 [62e3926e@gateway/web/freenode/ip.98.227.146.110] has quit [Quit: Page closed] | 03:59 | |
-!- sks [~sks@14.97.196.191] has joined #shogun | 06:31 | |
-!- zxtx [~zv@cpe-98-154-83-76.socal.res.rr.com] has joined #shogun | 06:49 | |
-!- zxtx [~zv@cpe-98-154-83-76.socal.res.rr.com] has quit [Ping timeout: 255 seconds] | 07:34 | |
-!- zxtx [~zv@cpe-98-154-83-76.socal.res.rr.com] has joined #shogun | 07:34 | |
-!- zxtx [~zv@cpe-98-154-83-76.socal.res.rr.com] has quit [Ping timeout: 264 seconds] | 07:40 | |
-!- zxtx [~zv@cpe-98-154-83-76.socal.res.rr.com] has joined #shogun | 07:45 | |
-!- zxtx [~zv@cpe-98-154-83-76.socal.res.rr.com] has quit [Read error: Operation timed out] | 07:50 | |
-!- zxtx [~zv@cpe-98-154-83-76.socal.res.rr.com] has joined #shogun | 07:50 | |
-!- vojtech [d975d865@gateway/web/freenode/ip.217.117.216.101] has joined #shogun | 09:30 | |
-!- zxtx [~zv@cpe-98-154-83-76.socal.res.rr.com] has quit [Ping timeout: 245 seconds] | 09:31 | |
-!- zxtx [~zv@cpe-75-83-151-252.socal.res.rr.com] has joined #shogun | 09:55 | |
-!- emrecelikten [~emre@213.153.222.106] has quit [Quit: Leaving.] | 09:57 | |
-!- hsyal_ [~hsyal@14.140.239.126] has joined #shogun | 09:58 | |
-!- gsomix [~gsomix@109.169.225.237] has joined #shogun | 10:03 | |
gsomix | good morning | 10:05 |
-!- hsyal_ [~hsyal@14.140.239.126] has quit [Ping timeout: 264 seconds] | 10:06 | |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun | 10:23 | |
-!- sks [~sks@14.97.196.191] has quit [Ping timeout: 244 seconds] | 11:56 | |
-!- blackburn [~blackburn@81.28.187.0] has joined #shogun | 12:00 | |
@sonney2k | morning gsomix | 12:01 |
@sonney2k | I hope you find your way through swig and python :) | 12:05 |
-!- vojtech [d975d865@gateway/web/freenode/ip.217.117.216.101] has quit [Quit: Page closed] | 12:30 | |
-!- hsyal [~hsyal@14.140.239.126] has joined #shogun | 12:39 | |
hsyal | hey n4nd0 , hows your gsoc going ?? | 12:40 |
n4nd0 | hsyal: hey! it is going fine, what about you? | 12:41 |
n4nd0 | still in amazon? | 12:41 |
hsyal | yeah.. | 12:41 |
n4nd0 | cool | 12:41 |
n4nd0 | what are you doing around there? | 12:41 |
hsyal | need your help | 12:41 |
hsyal | some android stuff | 12:41 |
n4nd0 | I am not mobile developer to tell the truth .. but tell me | 12:42 |
hsyal | no no.. I need some help on ML only | 12:42 |
n4nd0 | ah all right | 12:42 |
hsyal | I wrote "some android stuff " to answer your question | 12:42 |
n4nd0 | yeah, got it ;) | 12:42 |
hsyal | Actually I need to some to apply some regression stuff | 12:42 |
hsyal | on some data | 12:42 |
n4nd0 | ok | 12:43 |
hsyal | I need some algo which is non-linear, | 12:43 |
blackburn | KRR, SVR | 12:43 |
hsyal | hey blackburn , thanks... | 12:45 |
hsyal | can I extract parameters from those algos | 12:45 |
blackburn | :) | 12:45 |
blackburn | parameters like? | 12:45 |
hsyal | by parameters I mean theta's | 12:47 |
hsyal | generated in hypothesis function | 12:47 |
hsyal | Actually I am not in touch with ml, so forgot all jargon. | 12:47 |
blackburn | if you want to learn some model (like quadratic or so) you may use some hand-tailored solution I think | 12:48 |
blackburn | just formulate loss function and pass it to some optimizer | 12:49 |
hsyal | okay... | 12:49 |
hsyal | but it would take time | 12:49 |
blackburn | I am not sure about extracting parameters from alphas, n4nd0 are you? | 12:49 |
hsyal | Actually, I want to embed that hypothesis function inside android program. | 12:50 |
hsyal | So if have any other ideas please do share | 12:50 |
blackburn | http://docs.scipy.org/doc/scipy/reference/tutorial/optimize.html#least-square-fitting-leastsq | 12:51 |
n4nd0 | blackburn, hsyal I don't really know either | 12:51 |
blackburn | hsyal: it really depends on the model you use | 12:51 |
blackburn | in KRR/SVR with say gaussian kernel you have model that can't be described somehow else | 12:52 |
blackburn | only using alphas and feature vectors you had | 12:52 |
hsyal | blackburn, what about linear regression in shogun, liblinear n ol | 12:53 |
blackburn | hsyal: you said you want non-linear ;) | 12:53 |
hsyal | can they give me hypothesis function.. | 12:53 |
blackburn | hsyal: yes, linear | 12:53 |
blackburn | if you have much dimensions it would work | 12:54 |
hsyal | yeah, but I can compromise on that .. | 12:54 |
n4nd0 | hsyal, blackburn: now that I think of it, maybe GPs could be useful | 12:54 |
hsyal | GP ?? forgot what it was.. | 12:54 |
blackburn | hsyal: gaussian processes | 12:55 |
hsyal | ohhh. | 12:55 |
blackburn | hsyal: just try linear and if you are not satisfied with performance go for nonlinear | 12:56 |
hsyal | and one more thing do you have any piecewise linear regression algo in shogun | 12:56 |
blackburn | no | 12:56 |
blackburn | I don't know any to be honest | 12:56 |
hsyal | Is that any better than non-linear, actually i dont even know what it is | 12:56 |
hsyal | just read it in one paper | 12:57 |
hsyal | and btw hows your work going in gsoc , blackburn | 12:57 |
n4nd0 | see you later guys | 13:00 |
blackburn | hsyal: pretty fine | 13:00 |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: leaving] | 13:01 | |
hsyal | blackburn, gr8 , thanks for your help.. see you later. | 13:03 |
blackburn | hsyal: see you | 13:03 |
-!- hsyal [~hsyal@14.140.239.126] has quit [Quit: Leaving] | 13:06 | |
-!- blackburn [~blackburn@81.28.187.0] has quit [Quit: Leaving.] | 13:15 | |
-!- nietpiet [92322838@gateway/web/freenode/ip.146.50.40.56] has joined #shogun | 13:16 | |
-!- hsyal [~hsyal@14.140.239.126] has joined #shogun | 13:21 | |
hsyal | hey guys facing this error: snipt.org/vmia8 at compile time. help ..! | 13:26 |
-!- sks [~sks@115.246.146.87] has joined #shogun | 13:44 | |
wiking | alex ! | 14:01 |
hsyal | wiking, can you help me with this error : www.snipt.org/vmia8 | 14:03 |
wiking | oh man i wish i could but i'm overwhelmed with one thing now | 14:05 |
wiking | :( | 14:05 |
wiking | have you tried to do a make clean && ./configure && make ? | 14:06 |
wiking | maybe you have changed something in your repo ever since you did the last make&configure combo | 14:07 |
-!- emrecelikten [~emre@213.153.216.212] has joined #shogun | 14:26 | |
-!- emrecelikten [~emre@213.153.216.212] has quit [Client Quit] | 14:30 | |
-!- hsyal [~hsyal@14.140.239.126] has quit [Ping timeout: 264 seconds] | 14:39 | |
-!- hsyal [~hsyal@14.140.239.126] has joined #shogun | 14:54 | |
-!- hsyal_ [~hsyal@14.140.239.126] has joined #shogun | 14:54 | |
-!- sks [~sks@115.246.146.87] has quit [Ping timeout: 250 seconds] | 14:54 | |
-!- sks [~sks@14.97.177.73] has joined #shogun | 15:12 | |
-!- nietpiet [92322838@gateway/web/freenode/ip.146.50.40.56] has quit [Quit: Page closed] | 15:17 | |
-!- sks [~sks@14.97.177.73] has quit [Ping timeout: 245 seconds] | 15:39 | |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun | 15:51 | |
n4nd0 | :qa | 15:54 |
n4nd0 | ups haha | 15:54 |
n4nd0 | wrong tab change | 15:54 |
-!- vojtech [d975d865@gateway/web/freenode/ip.217.117.216.101] has joined #shogun | 15:55 | |
vojtech | n4nd0, Hi. Did you get my email with the document outlining the SO-SVM formulations? | 16:09 |
-!- pluskid [~pluskid@111.120.11.213] has joined #shogun | 16:11 | |
n4nd0 | vojtech: hey, sorry for the delay to answer | 16:15 |
n4nd0 | no I didn't get it :( | 16:15 |
n4nd0 | vojtech: what mail address did you use? | 16:15 |
vojtech | cool, then I sent it to someone else :( | 16:19 |
vojtech | I used nico.goernitz@tu-berlin.de | 16:19 |
n4nd0 | ok, that's nico address :) | 16:19 |
n4nd0 | vojtech: try with this one | 16:20 |
vojtech | which one ? | 16:21 |
n4nd0 | I just sent it to you in a private window on IRC | 16:22 |
vojtech | ok, I have it | 16:22 |
n4nd0 | cool | 16:22 |
n4nd0 | vojtech: ok, I have the document | 16:25 |
n4nd0 | thank you very very much :) | 16:25 |
vojtech | Might be cool to have table matching nickames and real names as I'm a bit lost ... but may be I'm the only one :) | 16:25 |
vojtech | BTW, aren't you based in Vilnius? | 16:25 |
n4nd0 | vojtech: mm no, Stockholm | 16:26 |
n4nd0 | why? | 16:26 |
vojtech | I though you are connected via server cameron.freenode.net [Vilnius, Lithuania, EU] but it seems to be the IRC server, right? | 16:27 |
vojtech | and I' going to Vilnius on Sunday ... | 16:28 |
n4nd0 | I have never been there | 16:32 |
n4nd0 | I don't know what server I am connected from :D | 16:32 |
vojtech | n4nd0, I just sent you v2 of the document as I found some typos. | 16:38 |
-!- hsyal_ [~hsyal@14.140.239.126] has quit [Ping timeout: 264 seconds] | 16:42 | |
-!- hsyal [~hsyal@14.140.239.126] has quit [Ping timeout: 264 seconds] | 16:42 | |
n4nd0 | vojtech: thank you | 16:44 |
vojtech | as I wrote it is just a draft. If you found some mistakes or missing information let me know. | 16:45 |
vojtech | I'll complete the document when I have some time or on request :) | 16:46 |
n4nd0 | ok, good idea :) | 16:47 |
n4nd0 | vojtech: I think I don't understand the meaning of the Equation (4) | 16:57 |
n4nd0 | vojtech: I see that part of it says that \Delta(y_i, h(x_i, w)) is less or equal than the \Delta(y_i, y) for any other y \in Y | 17:04 |
n4nd0 | but I think I am missing something else in it | 17:04 |
gsomix | sonney2k, around? | 17:05 |
gsomix | sonney2k, it works. | 17:07 |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Ping timeout: 264 seconds] | 17:12 | |
vojtech | n4nd0, what are you missing in it ? I think it is correct.. | 17:12 |
-!- gsomix [~gsomix@109.169.225.237] has quit [Read error: Connection reset by peer] | 17:13 | |
-!- gsomix [~gsomix@109.169.225.237] has joined #shogun | 17:13 | |
vojtech | n4nd0, the first line (i.e. the equality) just expresses the fact that the loss is paid for the labeling with the maximal scoring function | 17:13 |
vojtech | n4nd0, the second (i.e the inequality (4)) is obtained by replacing one of the maximization with only one element of it | 17:14 |
-!- alexlovesdata_ [82955843@gateway/web/freenode/ip.130.149.88.67] has joined #shogun | 17:15 | |
-!- pluskid [~pluskid@111.120.11.213] has quit [Ping timeout: 250 seconds] | 17:16 | |
-!- pluskid [~pluskid@202.130.113.141] has joined #shogun | 17:16 | |
-!- pluskid [~pluskid@202.130.113.141] has quit [Quit: Leaving] | 17:33 | |
-!- nietpit [92322838@gateway/web/freenode/ip.146.50.40.56] has joined #shogun | 17:35 | |
-!- nietpit [92322838@gateway/web/freenode/ip.146.50.40.56] has quit [Client Quit] | 17:35 | |
-!- nietpiet [92322838@gateway/web/freenode/ip.146.50.40.56] has joined #shogun | 17:37 | |
-!- gsomix [~gsomix@109.169.225.237] has quit [Quit: Ex-Chat] | 17:39 | |
-!- gsomix [~gsomix@109.169.225.237] has joined #shogun | 17:39 | |
-!- gsomix [~gsomix@109.169.225.237] has quit [Ping timeout: 244 seconds] | 17:58 | |
-!- gsomix [~gsomix@109.169.225.237] has joined #shogun | 18:23 | |
-!- vojtech [d975d865@gateway/web/freenode/ip.217.117.216.101] has quit [Quit: Page closed] | 18:25 | |
-!- blackburn [~blackburn@81.28.187.0] has joined #shogun | 18:30 | |
blackburn | wiking: what is problem with make? | 19:08 |
wiking | this is hilarious | 19:09 |
wiking | float64_t* w, cur_w; | 19:10 |
wiking | int32_t w_dim; | 19:10 |
wiking | memcpy (w, cur_w, w_dim); | 19:10 |
wiking | does not copy :( | 19:10 |
wiking | why? | 19:10 |
wiking | i have to do a for for (int i = 1; i < w_dim; ++i) w[i] = cur_w[i]; | 19:10 |
blackburn | why from 1? | 19:11 |
wiking | oh sorry | 19:11 |
wiking | so yeah i = 0 | 19:11 |
wiking | that's a typo | 19:11 |
blackburn | memcpy should work nicely hmm | 19:11 |
blackburn | ahh | 19:11 |
blackburn | memcpy(w,cur_w,sizeof(float64_t)*w_dim); | 19:11 |
blackburn | wiking: ^ | 19:12 |
blackburn | so probably it copied only 1/4 of needed data | 19:12 |
wiking | motherfuckeri'mblind | 19:12 |
wiking | yeey | 19:15 |
-!- blackburn [~blackburn@81.28.187.0] has left #shogun [] | 19:56 | |
-!- blackburn [~blackburn@81.28.187.0] has joined #shogun | 20:07 | |
-!- nietpiet [92322838@gateway/web/freenode/ip.146.50.40.56] has quit [Quit: Page closed] | 20:12 | |
alexlovesdata_ | memcpy tstststststs | 20:16 |
@sonney2k | gsomix, I heard you want to be the hero of the day :) | 20:18 |
gsomix | sonney2k, I am just a small man in a big world | 20:20 |
gsomix | sonney2k, http://pastebin.com/pkJ48gVQ | 20:20 |
gsomix | sonney2k, http://pastebin.com/hM8m0gUY spam.cpp | 20:21 |
gsomix | sonney2k, http://pastebin.com/fJB5umga spam.i | 20:21 |
gsomix | sorry for highlightnig | 20:21 |
blackburn | gsomix: I heard sonney2k loves it so feel free to highlight | 20:22 |
@sonney2k | blackburn, :P | 20:25 |
@sonney2k | gsomix, is the vector3d_get_item set_item stuff really needed? | 20:25 |
@sonney2k | I mean shouldn't swig do this? | 20:25 |
@sonney2k | ahh no you didn't overload operators so no | 20:26 |
@sonney2k | so you would have to do a.set* a.get* | 20:26 |
gsomix | sonney2k, internal_data is protected | 20:28 |
gsomix | not "swigged" | 20:28 |
gsomix | sonney2k, yep, I can overload get/set slots. | 20:29 |
gsomix | but I need some method to work with internal_data from interface file. | 20:30 |
@sonney2k | gsomix, but you have get_data_value and set_data_value | 20:31 |
gsomix | yes | 20:31 |
@sonney2k | so you can do this without these functions | 20:31 |
gsomix | awww, sorry. don't parse right. | 20:33 |
gsomix | I will try to overload operators. | 20:33 |
gsomix | get_item/set_item are not really needed | 20:33 |
-!- sks [~sks@14.97.18.251] has joined #shogun | 20:35 | |
@sonney2k | gsomix, it is not essential though | 20:35 |
gsomix | yes. I just have try overload sequence protocol functions. e.g. for slices | 20:35 |
gsomix | *tried | 20:36 |
gsomix | I also think that overload of old buffer protocol functions is needed. | 20:38 |
gsomix | e.g. for numpy.frombuffer(my_super_class) | 20:38 |
@sonney2k | gsomix, you mean for python2.X ? | 20:38 |
gsomix | 2.6< | 20:38 |
@sonney2k | gsomix, wait no | 20:38 |
@sonney2k | we require python 2.6 anyways | 20:38 |
gsomix | and for numpy.frombuffer(my_super_class) | 20:39 |
gsomix | it uses old buffer protocol | 20:39 |
gsomix | I don't know why. | 20:39 |
@sonney2k | gsomix, wait | 20:39 |
@sonney2k | there are 2 versions | 20:39 |
@sonney2k | the oldest one is the numpy array protocol | 20:40 |
@sonney2k | then there was the PEP with the buffer protocol for python 2.X | 20:40 |
@sonney2k | and then the newer python3 stuff | 20:40 |
@sonney2k | I think we should only support python2.x (x >=6) and python3 | 20:41 |
gsomix | ok | 20:42 |
@sonney2k | gsomix, btw which slot do you have to override to get slicing to work? | 20:42 |
@sonney2k | gsomix, btw when that stuff works reliably it will also fix the director dot feature issues - if you do buffer protocol support for the SG* datatypes | 20:43 |
@sonney2k | gsomix, anyways again *excellent* work | 20:44 |
@sonney2k | next steps would then be to implement that stuff for CDenseFeatures (aka RealFeatures, ByteFeatures etc in python) | 20:45 |
@sonney2k | then add some examples and then later do the same for the SG* data types and CustomKernel and CustomDistance (both based on SGMatrix) | 20:45 |
gsomix | >> to get slicing to work?| I didn't. Just enabled sequence protocol with getter/setter (vector3d_get_item). wait... hmm | 20:46 |
-!- sks [~sks@14.97.18.251] has quit [Ping timeout: 248 seconds] | 20:47 | |
gsomix | sonney2k, again, bad docs. I just need to overload sq_slice slot. | 20:48 |
gsomix | sonney2k, ok. | 20:49 |
gsomix | blackburn loves slices | 20:50 |
@sonney2k | gsomix, I do too | 20:50 |
gsomix | ok | 20:50 |
blackburn | everybody do | 20:50 |
@sonney2k | and it is soo easy to support them I think | 20:50 |
gsomix | yes! | 20:50 |
gsomix | cool stuff | 20:51 |
@sonney2k | gsomix, my suggestion is that you do this for densefeatures as next step then and then we find another challenge for you | 20:52 |
gsomix | ok | 20:53 |
@sonney2k | like these typemaps for model selection | 20:53 |
blackburn | to integrate shogun as kernel module | 20:53 |
@sonney2k | blackburn, wait the kernel as a shogun module ;-) | 20:54 |
@sonney2k | no but seriously, directors, buffer protocol support and these modelselection typemaps will increase useability by 1000% in python_modular | 20:55 |
@sonney2k | blackburn, I cannot currently think of other useability enhancements except for better docs and fixing the test suite | 20:56 |
@sonney2k | and better error messages | 20:56 |
blackburn | deb | 20:56 |
@sonney2k | daily .deb you mean | 20:56 |
@sonney2k | well yeah... | 20:56 |
@sonney2k | but that is relatively easy given that we have some packaging running already | 20:57 |
blackburn | no idea whether we really need it | 20:57 |
@sonney2k | blackburn, it certainly is very convenient to have though for newcomers | 20:58 |
blackburn | actually our users are not like vlc or so users | 20:58 |
@sonney2k | blackburn, yeah but they are also more on the theory side and struggle with install issues | 20:58 |
@sonney2k | gsomix, btw with all your reading swig docs - do you know whether it nowadays natively supports doxygen? | 20:59 |
@sonney2k | we have some hacks for that but just creating these docs in a python script takes 1 minute... | 21:00 |
@sonney2k | blackburn, btw I have one problem I could never solve: avoiding to copy memory between shogun feature objects and numpy matrices | 21:01 |
blackburn | no way | 21:01 |
gsomix | sonney2k, unfortunately I don't know. | 21:01 |
@sonney2k | gsomix, ok | 21:02 |
@sonney2k | blackburn, hehe I think this could be possible | 21:02 |
blackburn | how to let memory stay when np matrix is deleted& | 21:03 |
blackburn | ? | 21:03 |
@sonney2k | of course one has to specify the fortran flag to numpy matrices | 21:03 |
@sonney2k | otherwise it is not the same memory block | 21:03 |
@sonney2k | and no gaps etc | 21:03 |
@sonney2k | blackburn, well Py_INCREF(matrix) | 21:03 |
gsomix | blackburn, refcount for memory blocks? | 21:03 |
@sonney2k | and Py_DECREF in destructor | 21:04 |
@sonney2k | blackburn, non trivial business I guess but our hero of the day needs some serious challenges | 21:05 |
blackburn | sonney2k: I am just unsure we really need it | 21:06 |
@sonney2k | blackburn, if you have a 5G numpy matrix in memory that you want shogun to work with you wouldn't say this :) | 21:07 |
gsomix | sonney2k, btw, is memory copying in numpy.array(some_with_buffer)? | 21:08 |
blackburn | sonney2k: memory is cheap today :D | 21:08 |
@sonney2k | gsomix, I think so yes | 21:09 |
@sonney2k | blackburn, bad excuse - how much memory do you have in your notebook? | 21:09 |
blackburn | sonney2k: 8GB | 21:09 |
@sonney2k | so then a 7GB matrix it is :D | 21:09 |
@sonney2k | and don't tell me memory is cheap now! | 21:10 |
blackburn | sonney2k: I will use C++ then :D | 21:11 |
-!- alexlovesdata_ [82955843@gateway/web/freenode/ip.130.149.88.67] has quit [Quit: Page closed] | 21:15 | |
gsomix | >> I think so yes | hmmm. | 21:18 |
gsomix | sonney2k, I don't think so. | 21:20 |
gsomix | but I need to look into numpy code. | 21:21 |
@sonney2k | blackburn, haha | 21:25 |
@sonney2k | blackburn, ask cwidmer whether he would like that feature - *I* already know the answer :D | 21:26 |
blackburn | sonney2k: which feat? non-copy? | 21:26 |
@sonney2k | yes | 21:27 |
blackburn | I don't mind it while I don't have to do that :D | 21:27 |
@sonney2k | I think that should be some global flag or some boolean | 21:28 |
@sonney2k | (some boolean in each object) | 21:30 |
* sonney2k continues work on the buildbot | 21:31 | |
blackburn | sonney2k: why sometimes SG_ERROR is not being showed? | 21:32 |
@sonney2k | blackburn, that cannot be | 21:39 |
@sonney2k | blackburn, must be some corruption before | 21:39 |
gsomix | sonney2k, there is copy flag in numpy.array constructor. | 21:40 |
blackburn | sonney2k: yeah but that happens sometimes - I do not understand why | 21:40 |
gsomix | sonney2k, http://pastebin.com/Ax0ZhNtC | 21:41 |
@sonney2k | gsomix, so we could use the same... | 21:41 |
gsomix | yes | 21:41 |
gsomix | it just uses buffer protocol. | 21:41 |
@sonney2k | gsomix, when copy=True? | 21:42 |
@sonney2k | or always? | 21:42 |
gsomix | sonney2k, always. | 21:42 |
gsomix | but with copy=True memory is not copying | 21:43 |
gsomix | awww typo | 21:43 |
gsomix | *copy=False | 21:43 |
@sonney2k | gsomix, but how do they lock memory? | 21:46 |
@sonney2k | I mean they need to be sure it is not destroyed... | 21:46 |
gsomix | maybe Py_INCREF(exporter) in getbuffer function? | 21:47 |
gsomix | I really don't know | 21:47 |
gsomix | I need time to learn how this works | 21:48 |
@sonney2k | gsomix, problem with swig and objects is certainly that we don't currently know how to DECREF | 21:49 |
@sonney2k | just because we never store the original PyObject ptr | 21:50 |
@sonney2k | gsomix, yes you are right incref the exporter should do it | 21:59 |
@sonney2k | gsomix, but you need to decref it somehow - I hope that is done on decref'ing view? | 22:00 |
gsomix | sonney2k, PyBuffer_Release can decref exporter. | 22:03 |
gsomix | about bf_releasebuffer: | 22:04 |
gsomix | This function MUST NOT decrement view->obj, since that is done automatically in PyBuffer_Release() (this scheme is useful for breaking reference cycles). | 22:04 |
gsomix | PyBuffer_Release() is the interface for the consumer that wraps this function. | 22:05 |
CIA-18 | shogun: Michal Uricar master * r56632ab / (2 files): libbmrm inactive cutting plane removal - http://git.io/wMNAfg | 22:06 |
CIA-18 | shogun: Michal Uricar master * r5fa61fc / (4 files): DualLibQPBMSOSVM get_result method added + libbmrm update - http://git.io/bzgvAg | 22:06 |
CIA-18 | shogun: Soeren Sonnenburg master * r564295a / (4 files): Merge pull request #619 from uricamic/BM_SOL_EXAMPLE - http://git.io/HcoIpw | 22:06 |
CIA-18 | shogun: Michal Uricar master * ra90777e / (2 files): DualLibQPBMSOSVM BM parameters added. - http://git.io/pcpqzg | 22:06 |
blackburn | QPBMSOSVM BM?! | 22:07 |
blackburn | WTF :D | 22:07 |
shogun-buildbot | build #68 of deb1 - libshogun started, including [a90777ee041ed943619d1d10428c2b73236e40a3] | 22:07 |
shogun-buildbot | build #69 of deb1 - libshogun started, including [56632ab4ae0b4b00ce41e8faad36374ffaf76687, 5fa61fca2fdc2357437fff223c3b2a76b0597be4] | 22:07 |
shogun-buildbot | build #68 of deb1 - libshogun is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb1%20-%20libshogun/builds/68 blamelist: Michal Uricar <uricar.michal@gmail.com> | 22:08 |
shogun-buildbot | build #70 of deb1 - libshogun started, including [564295ab9c61f31d38587cd64212cb7587772925] | 22:08 |
shogun-buildbot | build #70 of deb1 - libshogun is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb1%20-%20libshogun/builds/70 | 22:09 |
shogun-buildbot | build #79 of deb3 - modular_interfaces started, including [564295ab9c61f31d38587cd64212cb7587772925] | 22:09 |
@sonney2k | blackburn, bundle method == BM | 22:14 |
blackburn | yeah | 22:14 |
blackburn | quadratic programming bundle method structured output support vector machine | 22:14 |
blackburn | I guess | 22:14 |
@sonney2k | blackburn, hmmhh QuadraticProgrammingBundleMethodStructuredOutputSupportVectorMachine :D | 22:15 |
@sonney2k | that's probably above the 80 char line width limit :D | 22:15 |
shogun-buildbot | build #69 of deb1 - libshogun is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb1%20-%20libshogun/builds/69 | 22:16 |
shogun-buildbot | build #78 of deb2 - static_interfaces started, including [564295ab9c61f31d38587cd64212cb7587772925] | 22:16 |
-!- sks [~sks@115.246.255.174] has joined #shogun | 22:21 | |
shogun-buildbot | build #78 of deb2 - static_interfaces is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb2%20-%20static_interfaces/builds/78 | 22:26 |
shogun-buildbot | build #80 of deb3 - modular_interfaces started, including [56632ab4ae0b4b00ce41e8faad36374ffaf76687, 5fa61fca2fdc2357437fff223c3b2a76b0597be4] | 22:26 |
-!- sks_ [~sks@115.247.10.56] has joined #shogun | 22:31 | |
gsomix | sonney2k, btw, http://pastebin.com/ZZVJ99vX | 22:31 |
gsomix | I'll try some experiment tomorrow. | 22:33 |
gsomix | make vector3d(some_with_buffer) constructor w/o copying | 22:33 |
-!- sks [~sks@115.246.255.174] has quit [Ping timeout: 252 seconds] | 22:33 | |
gsomix | after work on DenseFeatures | 22:33 |
@sonney2k | gsomix, cool | 22:36 |
@sonney2k | gsomix, we need to somehow ensure that we don't allow any kind of data modifications if we have copy=false | 22:37 |
@sonney2k | so preprocessors changing shape or even freeing memory -> blackburn HAHA | 22:37 |
@sonney2k | are a no go then | 22:38 |
-!- sks_ [~sks@115.247.10.56] has quit [Ping timeout: 252 seconds] | 22:38 | |
blackburn | sonney2k: why should I care about that? | 22:38 |
@sonney2k | blackburn, you will receive the crash reports - IIRC you do SG_FREE's of the matrix | 22:39 |
blackburn | sonney2k: only in dimredpreprocessor which is not used | 22:39 |
blackburn | converters ftw | 22:39 |
-!- sks_ [~sks@115.247.10.56] has joined #shogun | 22:41 | |
@sonney2k | blackburn, any comments about the cygwin buildbot? should we really try to run all the stuff we currently do also on cygwin? | 22:44 |
@sonney2k | I am a bit worried that the build bot might be too slow | 22:45 |
@sonney2k | (due to cygwin being slow) | 22:45 |
blackburn | sonney2k: no we shouldn't | 22:46 |
blackburn | :D | 22:46 |
@sonney2k | blackburn, so what then? | 22:46 |
@sonney2k | deb1 - deb3 -> cyg1 - cyg3 ? | 22:46 |
shogun-buildbot | build #79 of deb3 - modular_interfaces is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/79 | 22:47 |
shogun-buildbot | build #79 of deb2 - static_interfaces started, including [56632ab4ae0b4b00ce41e8faad36374ffaf76687, 5fa61fca2fdc2357437fff223c3b2a76b0597be4] | 22:47 |
@sonney2k | no nightly stuff? | 22:47 |
blackburn | I'd like to put 'if you are using windows' and a link to this song http://www.youtube.com/watch?v=p1BmY_HrAlw into disclaimer | 22:47 |
blackburn | sonney2k: did you get TU bot to wrk? | 22:48 |
@sonney2k | blackburn, it is running since yesterday | 22:50 |
@sonney2k | I now just compiled buildslave on the cygwin bot | 22:51 |
wiking | uristenmegbasztaaaaztaabudoszKURVA VILAGOT azt bassza meg | 22:51 |
blackburn | hungarian obscene lessons | 22:51 |
blackburn | kurva sounds bad for sure | 22:51 |
wiking | gsomix: clabels::get_num_labels isn't const? | 22:51 |
shogun-buildbot | build #79 of deb2 - static_interfaces is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb2%20-%20static_interfaces/builds/79 | 22:52 |
wiking | i mean it is not at the moment but i was just wondering if there's a reason for it that i'm too blind to realise why | 22:53 |
wiking | like the other memcpy story | 22:53 |
wiking | ... | 22:53 |
blackburn | wiking: no reason - there is a guy that doesn't like const | 22:53 |
wiking | ah ok | 22:53 |
wiking | just that i've got into a catch 22 situation | 22:53 |
blackburn | hah but actually | 22:53 |
blackburn | catch 22? | 22:53 |
gsomix | wiking, I dunno, lol | 22:53 |
wiking | blackburn: http://en.wikipedia.org/wiki/Catch-22 | 22:54 |
@sonney2k | blackburn, did const ever rescue you? | 22:54 |
blackburn | wiking: hah! I have been reading this book in russian before | 22:54 |
blackburn | didn't recognize though :( | 22:54 |
blackburn | sonney2k: my feets did a few times but const didn't for sure | 22:55 |
-!- nietpiet [92329056@gateway/web/freenode/ip.146.50.144.86] has joined #shogun | 22:55 | |
gsomix | wiking, unfortunately I did not have time to do all edits for const methods :( | 22:55 |
@sonney2k | wiking, if you think const helps you const all of the world! | 22:55 |
@sonney2k | gsomix, no worries there are certainly more useful tasks than a const triage | 22:56 |
blackburn | consthater | 22:56 |
@sonney2k | constantinople | 22:56 |
wiking | gsomix: i mean i really like the other changes, thnx a lot for that | 22:56 |
@sonney2k | constantinope | 22:56 |
blackburn | wiking: which changes? | 22:57 |
@sonney2k | ahhh | 22:57 |
* gsomix left to watch Futurama... | 22:58 | |
@sonney2k | swig on cygwin is 2.0.0 in the *experimental* distribution | 22:58 |
@sonney2k | gsomix, I watched them all | 22:58 |
blackburn | sonney2k: do you mind I add YES and NO as true and false synonyms somewhere? | 22:58 |
blackburn | :D | 22:58 |
@sonney2k | my biggest favorite is "Roswell that ends well" | 22:58 |
wiking | blackburn: the const changes in some of the apis | 22:59 |
blackburn | wiking: why do you care SO much? | 22:59 |
@sonney2k | gsomix, and I love the http://r33b.net/ | 22:59 |
wiking | blackburn: because trying to merge... and i've got a segfault atm | 22:59 |
blackburn | ?? | 23:00 |
wiking | unrelated actually :0 | 23:00 |
blackburn | because of const? | 23:00 |
blackburn | :D | 23:00 |
wiking | no that linking problem | 23:00 |
gsomix | sonney2k, \(^_^)/ I have not watched only 6 season. | 23:00 |
blackburn | ALL GLORY TO THE VALGRIND | 23:00 |
wiking | *was | 23:00 |
wiking | but now it's an actually segfault | 23:00 |
@sonney2k | gsomix, yeah you were too young :D | 23:00 |
@sonney2k | blackburn, lets do one thing with the cygwin buildbot | 23:01 |
@sonney2k | build libshogun! | 23:01 |
blackburn | sonney2k: makes sense | 23:02 |
@sonney2k | all the rest currently has problems | 23:02 |
blackburn | sonney2k: but I really think we should put F OFF for windows users :D | 23:02 |
@sonney2k | when swig 2.0.4 or greater is available we do python_modular | 23:02 |
blackburn | before I thought it is cool to support windows | 23:02 |
blackburn | now I do not | 23:02 |
* sonney2k tries to compile shogun on cygwin | 23:03 | |
wiking | doh | 23:04 |
@sonney2k | blackburn, failed on make depend | 23:04 |
@sonney2k | OM*G | 23:05 |
gsomix | blackburn, windows for dota2 only, kekeke | 23:05 |
blackburn | sonney2k: look at this guy he keeps saying thing about these shity games to get me nervous :D | 23:06 |
-!- sks_ [~sks@115.247.10.56] has quit [Ping timeout: 245 seconds] | 23:10 | |
shogun-buildbot | build #80 of deb3 - modular_interfaces is complete: Failure [failed test python_modular] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/80 blamelist: Michal Uricar <uricar.michal@gmail.com> | 23:17 |
@sonney2k | blackburn, I am unable to use find under cygwin | 23:28 |
@sonney2k | find shogun -name '*.templ' | 23:28 |
@sonney2k | doesnt' find a file | 23:28 |
@sonney2k | haha | 23:38 |
@sonney2k | naming clash | 23:38 |
@sonney2k | find from windows is taken :/ | 23:38 |
@sonney2k | no wonder | 23:38 |
blackburn | :D | 23:41 |
@sonney2k | blackburn, seen nietpiet's email | 23:42 |
@sonney2k | I think we have some major corruption going on here | 23:42 |
blackburn | sonney2k: cool | 23:43 |
-!- alexlovesdata [55b2212c@gateway/web/freenode/ip.85.178.33.44] has joined #shogun | 23:45 | |
-!- shogun-buildbot [~shogun-bu@7nn.de] has quit [Remote host closed the connection] | 23:49 | |
-!- shogun-buildbot [~shogun-bu@7nn.de] has joined #shogun | 23:52 | |
wiking | in DynamicObjectArray the get_num_elements returns 0 for me if there are no elements in it... although i've called the ctor with a given array size... is this intentional? | 23:59 |
@sonney2k | gsomix, ^ | 23:59 |
--- Log closed Sat Jul 07 00:00:17 2012 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!