| --- Log opened Wed Apr 10 00:00:29 2019 | ||
| -!- Moatman [~Moatman@pool-96-255-151-151.washdc.fios.verizon.net] has joined #shogun | 01:41 | |
| -!- TheMoatman [~Moatman@pool-96-255-151-151.washdc.fios.verizon.net] has joined #shogun | 02:05 | |
| -!- Moatman [~Moatman@pool-96-255-151-151.washdc.fios.verizon.net] has quit [Read error: Connection reset by peer] | 02:05 | |
| -!- wiking [~wiking@c-185-45-237-122.customer.ggaweb.ch] has joined #shogun | 04:21 | |
| -!- wiking [~wiking@c-185-45-237-122.customer.ggaweb.ch] has quit [Changing host] | 04:21 | |
| -!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 04:21 | |
| -!- mode/#shogun [+o wiking] by ChanServ | 04:21 | |
| -!- wiking [~wiking@huwico/staff/wiking] has quit [Ping timeout: 246 seconds] | 04:25 | |
| -!- Moatman [~Moatman@pool-96-255-151-151.washdc.fios.verizon.net] has joined #shogun | 05:02 | |
| -!- TheMoatman [~Moatman@pool-96-255-151-151.washdc.fios.verizon.net] has quit [Read error: Connection reset by peer] | 05:02 | |
| -!- Moatman [~Moatman@pool-96-255-151-151.washdc.fios.verizon.net] has quit [Remote host closed the connection] | 05:16 | |
| -!- Moatman [~Moatman@pool-96-255-151-151.washdc.fios.verizon.net] has joined #shogun | 06:16 | |
| -!- wiking [~wiking@c-185-45-237-122.customer.ggaweb.ch] has joined #shogun | 07:17 | |
| -!- wiking [~wiking@c-185-45-237-122.customer.ggaweb.ch] has quit [Changing host] | 07:17 | |
| -!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 07:17 | |
| -!- mode/#shogun [+o wiking] by ChanServ | 07:18 | |
| -!- Moatman [~Moatman@pool-96-255-151-151.washdc.fios.verizon.net] has quit [Remote host closed the connection] | 07:19 | |
| -!- wiking [~wiking@huwico/staff/wiking] has quit [Remote host closed the connection] | 07:38 | |
| -!- wiking_ [~wiking@c-185-45-237-122.customer.ggaweb.ch] has joined #shogun | 08:13 | |
| -!- wiking_ [~wiking@c-185-45-237-122.customer.ggaweb.ch] has quit [Remote host closed the connection] | 08:29 | |
| -!- wiking_ [~wiking@2001:67c:10ec:5784:8000::3ff] has joined #shogun | 08:54 | |
| -!- wiking_ [~wiking@2001:67c:10ec:5784:8000::3ff] has quit [Ping timeout: 250 seconds] | 08:58 | |
| -!- wiking_ [~wiking@2001:67c:10ec:5784:8000::3ff] has joined #shogun | 09:12 | |
| -!- wiking_ is now known as wiking | 09:19 | |
| -!- wiking [~wiking@2001:67c:10ec:5784:8000::3ff] has quit [Changing host] | 09:19 | |
| -!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 09:19 | |
| -!- mode/#shogun [+o wiking] by ChanServ | 09:19 | |
| -!- geektoni [c1cdd24a@gateway/web/freenode/ip.193.205.210.74] has joined #shogun | 09:44 | |
| geektoni | ping wiking | 09:45 |
|---|---|---|
| @wiking | pongi | 09:45 |
| @wiking | ok so i was just wondering about this | 09:45 |
| @wiking | lemme dig out | 09:45 |
| @wiking | ok so it's all about this crazy stuff https://github.com/shogun-toolbox/shogun/pull/4557/files#diff-1689b0bd3ceb28355cc3c05ed7ffe4f0R1135 | 09:47 |
| @wiking | i mean that dispatch code :) | 09:47 |
| @wiking | i've talked a bit about this Gil and i told him the same that this feels just wrong :) | 09:47 |
| @wiking | i mean that if else story | 09:47 |
| @wiking | especially that we have visitors for any :) | 09:48 |
| @wiking | so i was wondering whether we could actually change this into a base interface visitor | 09:48 |
| @wiking | if u see what i mean | 09:48 |
| geektoni | I see I see | 09:49 |
| geektoni | mmh so | 09:49 |
| geektoni | let me look a bit into it | 09:49 |
| @wiking | k | 09:50 |
| -!- gf712 [90520840@gateway/web/freenode/ip.144.82.8.64] has joined #shogun | 09:54 | |
| -!- wiking [~wiking@huwico/staff/wiking] has quit [Remote host closed the connection] | 10:16 | |
| -!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 10:41 | |
| -!- mode/#shogun [+o wiking] by ChanServ | 10:41 | |
| @wiking | gf712: https://github.com/shogun-toolbox/shogun/pull/4594 simplified | 11:01 |
| -!- gf712_ [90520840@gateway/web/freenode/ip.144.82.8.64] has joined #shogun | 11:12 | |
| gf712_ | wiking: cool having a look at the pr now | 11:12 |
| gf712_ | wiking: why do you use std::list? | 11:12 |
| @wiking | why not? | 11:12 |
| @wiking | :) | 11:12 |
| gf712_ | instead of unordered map? | 11:12 |
| @wiking | this is the only usecase for that struct atm | 11:12 |
| gf712_ | I guess, you don't need to look up | 11:13 |
| @wiking | and that will always have iterate through all elements | 11:13 |
| @wiking | :) | 11:13 |
| gf712_ | aight I see | 11:13 |
| -!- gf712 [90520840@gateway/web/freenode/ip.144.82.8.64] has quit [Ping timeout: 256 seconds] | 11:13 | |
| -!- gf712_ is now known as gf712 | 11:13 | |
| @wiking | return sgo_details::get_by_tag(this, name, std::move(sgo_details::GetByName())); | 11:16 |
| @wiking | does that std::move really needed? | 11:16 |
| gf712 | no | 11:16 |
| gf712 | it is bothering me | 11:16 |
| @wiking | me too | 11:16 |
| @wiking | i think that bs | 11:16 |
| @wiking | why was that actually added there? | 11:16 |
| gf712 | should be trivially copyable | 11:16 |
| @wiking | yes | 11:17 |
| gf712 | you need to ask Heiko | 11:17 |
| @wiking | and its a temp obj | 11:17 |
| gf712 | talking about move | 11:17 |
| @wiking | yes | 11:18 |
| gf712 | wiking: do you think it is good design to have a function that takes a r val ref of a type and one for pointers? | 11:18 |
| @wiking | motherfucker | 11:18 |
| @wiking | mmmm | 11:18 |
| gf712 | r val wouldn't copy | 11:18 |
| @wiking | i was wondering a lot about this | 11:18 |
| gf712 | and pointer would | 11:18 |
| @wiking | i mean copying a pointer | 11:18 |
| @wiking | is not a big deal right ?:) | 11:18 |
| gf712 | unless you don't need it anymore in that scope | 11:19 |
| @wiking | i mean it's just a 64 bit basetype | 11:19 |
| gf712 | and want to transfer ownership | 11:19 |
| gf712 | rather than copy it again | 11:19 |
| gf712 | this is for the grid search stuff | 11:19 |
| gf712 | I want an internal method that attaches a new node and copies it | 11:19 |
| gf712 | i.e. it comes from an interface | 11:19 |
| gf712 | and another one that just puts the node in the container without copy, because it is used from cpp | 11:20 |
| gf712 | I just don't know if it will be confusing for a future dev | 11:20 |
| @wiking | pushed a fix into develop | 11:20 |
| @wiking | :D | 11:20 |
| gf712 | nutter | 11:20 |
| gf712 | :D | 11:20 |
| gf712 | wiking: non virtual functions don't have to be any order in classes afaik | 11:27 |
| @wiking | i've got a compiler error | 11:28 |
| @wiking | :) | 11:28 |
| @wiking | not a warning | 11:28 |
| @wiking | :> | 11:28 |
| gf712 | so weird | 11:28 |
| gf712 | what compiler? | 11:28 |
| @wiking | default clang on osx | 11:29 |
| @wiking | Apple LLVM version 10.0.1 (clang-1001.0.46.3) | 11:29 |
| gf712 | weird that CI didn't complain | 11:29 |
| @wiking | indeed | 11:29 |
| gf712 | I don't see why order matters though | 11:30 |
| gf712 | but oh well | 11:30 |
| @wiking | i'm trying to finish deadbeef | 11:30 |
| @wiking | did a rebase | 11:30 |
| @wiking | and boom | 11:30 |
| @wiking | suddenly DynamicObjArray is like WTF | 11:30 |
| gf712 | ah cool, you putting js aside then? | 11:30 |
| @wiking | just for today | 11:31 |
| @wiking | i wanna finish this stuff | 11:31 |
| gf712 | yea | 11:31 |
| @wiking | as there's a lot of hanging prs | 11:31 |
| gf712 | btw can merge the to_string | 11:31 |
| gf712 | just commented on the includes | 11:31 |
| @wiking | lets see what CI says | 11:31 |
| gf712 | but don't matter | 11:31 |
| @wiking | i've fixed those | 11:31 |
| @wiking | :) | 11:31 |
| gf712 | ah ok! | 11:31 |
| @wiking | already pushed thnx for the comment | 11:31 |
| gf712 | need more more jobs on CI :D | 11:32 |
| @wiking | ha | 11:32 |
| @wiking | SG_FORCED_INLINE auto get_element_safe(size_t index) const | 11:32 |
| @wiking | this is why i think i had trouble | 11:32 |
| @wiking | :) | 11:32 |
| gf712 | what is causing the issue? | 11:34 |
| gf712 | not declared? | 11:34 |
| @wiking | auto? | 11:34 |
| gf712 | why is that an issue? should be able to deduce the type | 11:35 |
| @wiking | should be | 11:35 |
| @wiking | otherwise i have no clue | 11:35 |
| @wiking | why it failed | 11:35 |
| gf712 | is it working now then? | 11:36 |
| @wiking | with changing order yes | 11:37 |
| @wiking | :D | 11:37 |
| @wiking | man the fucking windows pr needs to be fixed as well | 11:37 |
| @wiking | :))) | 11:37 |
| @wiking | but yeah that depends on deadbeef | 11:37 |
| gf712 | oh the thing the export? | 11:38 |
| @wiking | yes | 11:38 |
| @wiking | so that actually there is a working dll on windows | 11:38 |
| gf712 | lol | 11:38 |
| gf712 | yea | 11:38 |
| gf712 | I actually now notice how few projects actually put the export | 11:38 |
| gf712 | and claim to work on windows | 11:38 |
| gf712 | wiking: were you still having issues with the deserialisation on windows for the tests? | 11:39 |
| @wiking | yeah windozer needs it | 11:39 |
| @wiking | yes | 11:39 |
| @wiking | that's why i revived the deadbeef pr | 11:39 |
| @wiking | from trash | 11:39 |
| @wiking | :D | 11:39 |
| @wiking | because i've got tired | 11:39 |
| @wiking | that i'm not seeing what the fuck is happening | 11:40 |
| @wiking | and didn't want to patch the old shit | 11:40 |
| @wiking | this way i should get a proper error why that fucking file is missing | 11:40 |
| @wiking | :D | 11:40 |
| @wiking | or that it fails | 11:40 |
| gf712 | haha | 11:41 |
| @wiking | need to figure out | 11:41 |
| @wiking | what utility function shoudl i add | 11:42 |
| @wiking | for serialization in SWIG | 11:42 |
| @wiking | as none of the fs stuff is SGobject | 11:42 |
| @wiking | and i would like to keep it like that | 11:42 |
| @wiking | and i have a branch locally | 11:44 |
| @wiking | that drops the fucking init.h story | 11:44 |
| @wiking | buuuuuut | 11:44 |
| @wiking | that actually is blocked by SGIO crazyness | 11:44 |
| @wiking | the MACROS are just crazy | 11:44 |
| @wiking | imo we should do something with SGIO | 11:44 |
| @wiking | i mean the whole story of formatting etc | 11:45 |
| @wiking | it's just craaazy | 11:45 |
| @wiking | i'm for spdlog-ing everything | 11:45 |
| @wiking | its nicely extensible with various sinks etc | 11:46 |
| @wiking | hehehe | 11:47 |
| @wiking | super! | 11:47 |
| @wiking | 197: [ RUN ] SparseFeaturesTest.serialization | 11:47 |
| @wiking | 197: unknown file: Failure | 11:47 |
| @wiking | 197: C++ exception with description "Out of memory error, tried to allocate -8270522240 bytes using malloc." thrown in the test body. | 11:47 |
| gf712 | wtf? | 11:48 |
| gf712 | makes no sense... | 11:48 |
| gf712 | spdlog-ing? | 11:48 |
| @wiking | gf712: https://github.com/gabime/spdlog | 11:49 |
| gf712 | ah looks cool | 11:50 |
| gf712 | so you can choose the output buffer? | 11:51 |
| @wiking | yes | 11:52 |
| @wiking | and of course you can write your own output buffer | 11:55 |
| @wiking | i once wrote a sink for slack | 11:55 |
| @wiking | was working nicely | 11:55 |
| @wiking | ok should be done... although its ugly | 11:58 |
| @wiking | pushing deadbeef | 11:58 |
| @wiking | unit tests should pass | 11:58 |
| gf712 | noiiccceeee | 11:58 |
| gf712 | who's reviewing it? :D | 11:58 |
| @wiking | nobody it seems | 11:58 |
| @wiking | :D | 11:58 |
| @wiking | but meta integration data should be still updated | 11:59 |
| @wiking | and swig story is still missing | 11:59 |
| @wiking | so still not mergable | 11:59 |
| gf712 | I tried but there's so much :D | 12:00 |
| gf712 | yea, I guess can discuss with Heiko? | 12:00 |
| gf712 | he said he would be around today | 12:00 |
| @wiking | yeah he told me the same as i cannot access gsoc dashboard | 12:01 |
| @wiking | and we should now have all the applications in | 12:01 |
| gf712 | oh damn, machine evaluation is leaking so much | 12:01 |
| @wiking | anyhow i'm off for lunch | 12:01 |
| gf712 | aight | 12:01 |
| gf712 | btw GSoC application deadline was extended | 12:01 |
| @wiking | really? | 12:02 |
| gf712 | yea, because of spring break | 12:02 |
| gf712 | pretty sure | 12:02 |
| @wiking | https://developers.google.com/open-source/gsoc/timeline | 12:02 |
| @wiking | April 9 18:00 UTC Student application deadline | 12:02 |
| @wiking | i've seen some mails about this | 12:02 |
| -!- HeikoS [~heiko@82-132-215-5.dab.02.net] has joined #shogun | 12:02 | |
| -!- mode/#shogun [+o HeikoS] by ChanServ | 12:02 | |
| @wiking | i followed it till the moment when gsoc admin wrote that it was working till now | 12:03 |
| @wiking | so why change | 12:03 |
| @wiking | D | 12:03 |
| @wiking | : | 12:03 |
| @wiking | :> | 12:03 |
| gf712 | ah sorry | 12:03 |
| gf712 | I thought they did | 12:03 |
| @wiking | i mean they have a point | 12:03 |
| gf712 | but was just someone complaining | 12:03 |
| @wiking | it was like this since forever | 12:03 |
| @wiking | yeah you always have a newcomer | 12:03 |
| @wiking | :D | 12:03 |
| @wiking | ok i'm out for 30ish | 12:03 |
| @wiking | ttyl | 12:03 |
| gf712 | ttyl | 12:04 |
| @HeikoS | yo! | 12:20 |
| gf712 | HeikoS: hey | 12:20 |
| gf712 | I am going through the emails you just sent | 12:21 |
| @HeikoS | cool | 12:21 |
| gf712 | wiking should be back soon | 12:21 |
| @HeikoS | good to put some thoughts into that | 12:21 |
| @HeikoS | distributed multi-language model selection might fly | 12:21 |
| gf712 | yea, I think it could be good | 12:21 |
| gf712 | and get something more long term and more deliverables | 12:21 |
| gf712 | and more people.. | 12:22 |
| gf712 | its a shame about the cohort though | 12:22 |
| gf712 | it would've been cool collaborating with other devs and exchange ideas | 12:22 |
| @HeikoS | yes we could do multiple work packages | 12:33 |
| @HeikoS | well you saw the email, that they want individual projects send proposals | 12:33 |
| gf712 | HeikoS: btw about the grid search, I have something more or less working | 12:55 |
| gf712 | just some odd refcounting going on I think | 12:55 |
| @HeikoS | gf712: cool | 13:00 |
| @HeikoS | what shall I look at? | 13:00 |
| @HeikoS | About to go for lunch, but will come back to pr reviews later | 13:02 |
| -!- HeikoS [~heiko@82-132-215-5.dab.02.net] has quit [Quit: Leaving.] | 13:03 | |
| -!- gf712 [90520840@gateway/web/freenode/ip.144.82.8.64] has quit [Ping timeout: 256 seconds] | 13:09 | |
| -!- gf712 [90520840@gateway/web/freenode/ip.144.82.8.64] has joined #shogun | 13:10 | |
| @wiking | uff | 13:18 |
| @wiking | 10 proposals :< | 13:19 |
| @wiking | rcurtin: how is your proposal count ? :) | 13:19 |
| @wiking | gf712: ping? | 13:20 |
| @wiking | do you get what is wrattler? :D | 13:20 |
| gf712 | wiking: i am not sure, might be confusing with something else, but I thinks its a notebook that support multiple languages | 13:24 |
| gf712 | so you write a cell in python, another in R | 13:24 |
| @wiking | yes | 13:24 |
| gf712 | and types get converted in between | 13:24 |
| @wiking | ok i've already wrote my 2p about it | 13:24 |
| @wiking | we coudl do something there for sure | 13:25 |
| gf712 | I have never tried it | 13:26 |
| gf712 | so I am not sure how it works | 13:26 |
| gf712 | or if it's stabble | 13:26 |
| gf712 | but can give it a go | 13:26 |
| @wiking | dunno why do they wanna distinguish themselves from iodide.io | 13:26 |
| @wiking | or not why but how | 13:27 |
| @wiking | :) | 13:27 |
| gf712 | I guess it has more languages? | 13:28 |
| gf712 | not that it really matters tbh | 13:28 |
| @wiking | btw gf712 4594 is merged | 13:28 |
| @wiking | gf712: iodide has py :) | 13:29 |
| gf712 | yup, just saw | 13:29 |
| gf712 | I'll rebase my code | 13:29 |
| gf712 | thanks! | 13:29 |
| gf712 | wiking: btw do all meta examples leak? | 13:29 |
| @wiking | hehehe could be :) | 13:30 |
| @wiking | if so we should fixor | 13:30 |
| gf712 | because valgrind aint happy | 13:30 |
| @wiking | yeah | 13:30 |
| @wiking | i remember we had this once | 13:30 |
| @wiking | i mean discussion about why and how | 13:30 |
| @wiking | things leak there | 13:30 |
| @wiking | but dunno what was the consensus | 13:30 |
| gf712 | and I am having a hard time distinguishing what is baseline leak | 13:30 |
| gf712 | and what is my code | 13:30 |
| @wiking | hehehe i think something around some is happening in a weird way :( | 13:30 |
| gf712 | the issue is that nothing is unrefed in the end | 13:31 |
| @wiking | ok now i have to run again | 13:31 |
| @wiking | gf712: yeah but one would hope that things are wrapped in some | 13:31 |
| @wiking | so that should be auto released | 13:31 |
| gf712 | valgrind says that the stuff allocated in factories is not getting delted | 13:31 |
| gf712 | deleted | 13:31 |
| gf712 | wiking: factory increments the counter | 13:42 |
| gf712 | and then so does wrap | 13:42 |
| gf712 | so I think one of them has to go no? | 13:42 |
| gf712 | because at the end unref is called only once by ~Some | 13:43 |
| gf712 | wiking: removing one of the ref increments solves memory issues | 13:55 |
| @wiking | here | 14:04 |
| @wiking | sorry | 14:04 |
| @wiking | gf712: ok so | 14:05 |
| @wiking | i mean the story is that factory should ideally return a some wrapped obj | 14:05 |
| @wiking | gf712: and actually we should have somes passed around in the codebase | 14:06 |
| @wiking | buuuut | 14:06 |
| @wiking | that has some swig problematics | 14:06 |
| @wiking | :) | 14:06 |
| @wiking | so for the time being | 14:06 |
| @wiking | factory shoudl return a ptr | 14:06 |
| @wiking | but yeah you are right | 14:06 |
| @wiking | in my opinion i would not have the SG_REF within the factory | 14:06 |
| @wiking | dunno how that will effect SWIG | 14:06 |
| @wiking | but it should be fine imo | 14:07 |
| @wiking | we shouldn't change the wrap behaviour of some imo | 14:07 |
| @wiking | or? | 14:07 |
| gf712 | but surely some should take ownership no? | 14:08 |
| @wiking | yes | 14:08 |
| gf712 | like a smart pointer | 14:08 |
| @wiking | yep but it currently does no? | 14:08 |
| @wiking | or you mean you wanna move the ptr ? :) | 14:08 |
| gf712 | well it does | 14:08 |
| gf712 | but that should reset the counter | 14:08 |
| gf712 | to 1 | 14:08 |
| @wiking | mmm | 14:08 |
| gf712 | rather than increment | 14:08 |
| @wiking | well that's a different type of smart pointer | 14:08 |
| @wiking | some is invasive :) | 14:08 |
| gf712 | but isn't that what a smart pointer does? | 14:09 |
| @wiking | i mean if currently | 14:09 |
| @wiking | no not necessarily | 14:09 |
| gf712 | it assumes it is the only owner | 14:09 |
| @wiking | no | 14:09 |
| @wiking | it can be anything | 14:09 |
| @wiking | :D | 14:09 |
| gf712 | and deletes when the refcount is 0 | 14:09 |
| @wiking | liek now we do it | 14:09 |
| @wiking | :> | 14:09 |
| @wiking | it is ANYTHING | 14:09 |
| @wiking | :) | 14:09 |
| gf712 | but then you leak | 14:09 |
| @wiking | nono | 14:09 |
| @wiking | i mean now some | 14:09 |
| @wiking | is there but since we have native poiinters in the codebase | 14:10 |
| @wiking | if we would start reseting ctr to 1 | 14:10 |
| @wiking | then things would totally crash | 14:10 |
| @wiking | just do a git grep SG_UNREF | 14:10 |
| @wiking | and u get what i mean | 14:10 |
| @wiking | since there's many places where we do manual sg_ref/unref | 14:10 |
| @wiking | but then sometimes we mix some in | 14:10 |
| @wiking | in this scenario | 14:10 |
| @wiking | if some resets the ctr to 1 | 14:11 |
| @wiking | then we are going to crash after every 2 calls :) | 14:11 |
| gf712 | hmmm ok | 14:11 |
| @wiking | what you are saying is true | 14:11 |
| @wiking | for a shared_ptr | 14:11 |
| gf712 | it's just confusing | 14:11 |
| @wiking | but some is an invasive smart ptr | 14:11 |
| gf712 | what about wrap does a move? | 14:11 |
| gf712 | and resets? | 14:11 |
| @wiking | and for the time being we have to live with this | 14:12 |
| gf712 | but some constructor doesn't | 14:12 |
| @wiking | until we get rid of native potiners | 14:12 |
| @wiking | *pointers | 14:12 |
| @wiking | yeah its a valid way to look at things but then again | 14:12 |
| @wiking | git grep wrap | 14:12 |
| @wiking | :) | 14:12 |
| @wiking | you'll see what i mean | 14:12 |
| gf712 | yup just did :( | 14:12 |
| @wiking | so i would just say | 14:13 |
| gf712 | BUT | 14:13 |
| @wiking | remove SG_REF from factories | 14:13 |
| gf712 | it's usually fine | 14:13 |
| gf712 | just need to change a couple of things | 14:13 |
| gf712 | hmmm ok | 14:13 |
| @wiking | and that should get thing rolling | 14:13 |
| @wiking | as anyways | 14:13 |
| @wiking | the output of factories | 14:13 |
| @wiking | will be taken by something | 14:13 |
| @wiking | right? | 14:13 |
| gf712 | yea, I guess so? | 14:13 |
| @wiking | so then | 14:13 |
| @wiking | they'll either do SG_REF | 14:14 |
| @wiking | or they'll do wrap | 14:14 |
| @wiking | :) | 14:14 |
| gf712 | true | 14:14 |
| @wiking | factory will just get you the obj | 14:14 |
| @wiking | and then you see what to do with it | 14:14 |
| @wiking | and where and how | 14:14 |
| gf712 | I just don't like the limbo of you have control or you don't and it alternates | 14:14 |
| @wiking | well | 14:14 |
| gf712 | it's like python API | 14:14 |
| @wiking | welcome to hell | 14:14 |
| @wiking | :> | 14:14 |
| gf712 | new stuff is refcounted | 14:14 |
| @wiking | but yeah i mean invasive smart ptr | 14:14 |
| gf712 | old stuff you have to do it | 14:14 |
| @wiking | is not our invention ;) | 14:15 |
| gf712 | yea I guess | 14:15 |
| gf712 | so one day there will be just smart pointers? | 14:15 |
| @wiking | for example embind (webassembly, has even a typedef for this) | 14:15 |
| @wiking | gf712: yeah i mean it woudl be great not to have those SG_REF/UNREF pairs | 14:15 |
| @wiking | in the codebase | 14:15 |
| @wiking | :) | 14:15 |
| @wiking | currently we have it because of SWIG | 14:16 |
| @wiking | we could totally switch to shared/unique_ptr or some/unique | 14:16 |
| gf712 | wiking: btw heap allocated objects without some in meta examples | 14:16 |
| gf712 | how do you think it should be done? | 14:16 |
| gf712 | objects that aren't sgobject | 14:16 |
| @wiking | mmm | 14:16 |
| @wiking | smart ptr | 14:16 |
| @wiking | if it's c++ | 14:16 |
| @wiking | :) | 14:16 |
| @wiking | if something else | 14:16 |
| @wiking | then we are having a problem | 14:17 |
| gf712 | shared_ptr | 14:17 |
| @wiking | :))) | 14:17 |
| gf712 | unique? | 14:17 |
| @wiking | whatever works | 14:17 |
| @wiking | but see my point | 14:17 |
| @wiking | that in swig | 14:17 |
| @wiking | this will be a problem | 14:17 |
| gf712 | yup | 14:17 |
| @wiking | doesn't matter if its smartptr or not | 14:17 |
| @wiking | that's why we have SG_REF/UNREF in the first place | 14:17 |
| @wiking | as in swig we define | 14:17 |
| @wiking | that ref() unref() is the thing to maintain reference count | 14:17 |
| gf712 | btw how does it work with %newobject? | 14:18 |
| @wiking | https://github.com/shogun-toolbox/shogun/blob/develop/src/interfaces/swig/SGBase.i#L12 | 14:18 |
| @wiking | gf712: that hasn't been figured out imo | 14:18 |
| gf712 | because I am using it | 14:18 |
| gf712 | with gridsearch | 14:18 |
| gf712 | but haven't figured out if python for example is leaking | 14:18 |
| @wiking | yeah | 14:19 |
| gf712 | I don't even know how to check it..? | 14:19 |
| @wiking | "If the user doesn't specify the "ref/unref" feature for a type, SWIG will produce code equivalent to defining these features:" | 14:20 |
| @wiking | but that's the problem | 14:20 |
| @wiking | currently we do define it | 14:20 |
| @wiking | right? :) | 14:20 |
| @wiking | this is the default | 14:20 |
| @wiking | %feature("ref") "" | 14:20 |
| @wiking | %feature("unref") "delete $this;" | 14:20 |
| @wiking | which would solve your problem right? :) | 14:20 |
| gf712 | ah for a type | 14:20 |
| gf712 | ok, makes sense | 14:21 |
| @wiking | but i dont know | 14:21 |
| @wiking | if you can specify this | 14:21 |
| @wiking | type wise | 14:21 |
| @wiking | :) | 14:21 |
| @wiking | or this is for the whole module | 14:21 |
| @wiking | imo module | 14:21 |
| @wiking | then you should look into definining a separate swig module | 14:21 |
| @wiking | gf712: read this section | 14:21 |
| @wiking | http://www.swig.org/Doc3.0/SWIGPlus.html#SWIGPlus_ref_unref | 14:21 |
| @wiking | so | 14:22 |
| @wiking | gf712: | 14:22 |
| @wiking | The %newobject feature is designed to indicate to the target language that it should take ownership of the returned object. When used in conjunction with a type that has the "ref" feature associated with it, it additionally emits the code in the "ref" feature into the C++ wrapper. Consider wrapping the following factory function in addition to the above: | 14:22 |
| @wiking | i think based on this | 14:22 |
| @wiking | i'm pretty sure that we dont need SG_REF within the factory code right? :) | 14:22 |
| @wiking | ok lemme try | 14:22 |
| gf712 | hmmm I guess so? | 14:23 |
| gf712 | how do you test for mem leak in a interface language? | 14:23 |
| @wiking | just a sec | 14:27 |
| gf712 | started a pr to see what ci says :p | 14:28 |
| @wiking | where? | 14:29 |
| @wiking | hahah | 14:29 |
| @wiking | w8 | 14:29 |
| @wiking | i'm getting you a much tricker pr | 14:29 |
| @wiking | ;) | 14:29 |
| gf712 | ah ok | 14:30 |
| gf712 | cancel the build if you want then | 14:30 |
| @wiking | githjub is craaazy | 14:33 |
| @wiking | :D | 14:33 |
| gf712 | what do you mean? | 14:34 |
| @wiking | ok something weird happend | 14:34 |
| @wiking | but here https://github.com/shogun-toolbox/shogun/pull/4604 | 14:34 |
| @wiking | ok i gotta run now | 14:35 |
| @wiking | but this should work | 14:35 |
| @wiking | apart from the c++ part maybe | 14:35 |
| gf712 | ah good! | 14:35 |
| @wiking | on swig level this should be the correct way | 14:35 |
| -!- wiking [~wiking@huwico/staff/wiking] has quit [Remote host closed the connection] | 14:35 | |
| gf712 | let's see | 14:35 |
| -!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 14:36 | |
| -!- mode/#shogun [+o wiking] by ChanServ | 14:36 | |
| @wiking | based on swig docs | 14:36 |
| @wiking | as newobj shoudl take care of calling ref() | 14:36 |
| @wiking | and i c++ we do anyways wrap | 14:36 |
| @wiking | but now i really gotta run | 14:37 |
| @wiking | ttyl | 14:37 |
| -!- wiking [~wiking@huwico/staff/wiking] has quit [Remote host closed the connection] | 14:37 | |
| -!- wiking [~wiking@178.197.225.168] has joined #shogun | 14:53 | |
| -!- wiking [~wiking@178.197.225.168] has quit [Changing host] | 14:53 | |
| -!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 14:53 | |
| -!- mode/#shogun [+o wiking] by ChanServ | 14:53 | |
| @wiking | gf712: sorry mate to close your pr | 14:56 |
| @wiking | :) | 14:56 |
| gf712 | haha | 14:56 |
| gf712 | cancel the build too | 14:56 |
| @wiking | done | 14:57 |
| gf712 | coolio | 15:03 |
| gf712 | so how do you check mem leaks in interface languages? | 15:04 |
| gf712 | I just use cpp and hope that is is representative... | 15:04 |
| -!- HeikoS [~heiko@82-132-215-5.dab.02.net] has joined #shogun | 15:05 | |
| -!- mode/#shogun [+o HeikoS] by ChanServ | 15:05 | |
| @wiking | HeikoS: yo | 15:05 |
| @HeikoS | yo | 15:11 |
| @HeikoS | w | 15:11 |
| @HeikoS | wiking: | 15:11 |
| @wiking | ho | 15:11 |
| @HeikoS | saw the bunch of emails re turing=? | 15:11 |
| @wiking | so i wrote some things about the wrattler story | 15:11 |
| @HeikoS | ah cool | 15:12 |
| @wiking | we should meet them | 15:12 |
| @HeikoS | I am not too sure about this | 15:12 |
| @HeikoS | wrattler in general, but might be worth a shot | 15:12 |
| @wiking | do we know about the proposal deadline? | 15:12 |
| @HeikoS | yes end of april something | 15:12 |
| @HeikoS | soon | 15:12 |
| @wiking | if there's such thing | 15:12 |
| @wiking | ok | 15:12 |
| @wiking | so | 15:12 |
| @wiking | then let's meet wrattler | 15:12 |
| @wiking | imo that's the best | 15:12 |
| @HeikoS | ill forward | 15:12 |
| @wiking | asap | 15:12 |
| @HeikoS | the thing is that the colab stuff | 15:12 |
| @wiking | so either this week still | 15:12 |
| @wiking | or early next week | 15:12 |
| @wiking | before easter | 15:12 |
| @HeikoS | on the ws they said that it shouldnt be between tools | 15:12 |
| @wiking | hehhe ok seems now its shifted | 15:13 |
| @HeikoS | but rather between tools and other projects | 15:13 |
| @wiking | let's adapt :D | 15:13 |
| @HeikoS | so you see | 15:13 |
| @HeikoS | this is why they sent the name list | 15:13 |
| @HeikoS | best would be to get someones name on the proposal | 15:13 |
| @wiking | i mean do we have another idea with whom we could work with? | 15:13 |
| @HeikoS | from another project | 15:13 |
| @wiking | ah ok | 15:13 |
| @HeikoS | i know a few of those names | 15:13 |
| @HeikoS | but also maybe you have an idea | 15:13 |
| @HeikoS | how to phrase stuff | 15:13 |
| @wiking | anybody from here | 15:14 |
| @wiking | Jan Lowe AI for Science Munro - Investigating the value of machine | 15:14 |
| @wiking | learning in predicting gene function | 15:14 |
| @wiking | ? | 15:14 |
| @HeikoS | yeah thought that | 15:14 |
| @HeikoS | because we have string kernels | 15:14 |
| @HeikoS | and citations | 15:14 |
| @wiking | Mark Birkin Digital Twins - Urban A Machine learning approach to | 15:14 |
| @wiking | understanding the disease trajectories of atrial fibrillation | 15:14 |
| @HeikoS | I want to prepare an intro email to a few names, suggesting what we do and then whether they would be keen to co-plan | 15:14 |
| @HeikoS | as note | 15:14 |
| @wiking | Spiros Denaxas Health Interpretable disease prediction models | 15:14 |
| @wiking | using supervised learning | 15:14 |
| @HeikoS | james said that other projects should be involved in the planning of the projects ... so I think if we have a name and some intention that should be enough for now | 15:15 |
| @wiking | these things could be something | 15:15 |
| @wiking | re: wrattler: my biggest concern is that seems like a redundant project of iodide | 15:16 |
| @wiking | but whatever | 15:16 |
| @wiking | lets try to talk | 15:16 |
| @wiking | anyhow any of the above mentioned we coudl work with | 15:16 |
| @HeikoS | yep I agree on wrattler | 15:18 |
| @HeikoS | seems like reinventing things | 15:18 |
| @wiking | anyhow... lets get this rollinng | 15:18 |
| @HeikoS | but we dont care about that as long as it helps us | 15:18 |
| @wiking | man n is stuck on my laptop | 15:18 |
| @wiking | :D | 15:18 |
| @HeikoS | lol | 15:18 |
| @wiking | this macbook is a piece of junk i must say | 15:18 |
| @wiking | thinkning about switching to x1 carbon | 15:19 |
| @HeikoS | thinkpad <3 | 15:19 |
| @HeikoS | I really like mine | 15:19 |
| @wiking | yeah.... i had a t42 | 15:19 |
| @wiking | but that was still ibm era | 15:19 |
| @HeikoS | didnt like that last few ts, but the x1 is quite nice | 15:19 |
| @wiking | u have a x1 carbon? | 15:19 |
| @HeikoS | yes | 15:19 |
| @wiking | cool... hdr and all? :D | 15:19 |
| @HeikoS | for 2-3 years now | 15:19 |
| @HeikoS | and cant really complain | 15:19 |
| @HeikoS | only about ubuntu | 15:19 |
| @wiking | touch part works under linux? | 15:19 |
| @HeikoS | the new one is good | 15:20 |
| @wiking | i mean it has a touch screen function right? | 15:20 |
| @HeikoS | the old one (also in the t) was shit | 15:20 |
| @HeikoS | no i have without | 15:20 |
| @HeikoS | optional | 15:20 |
| @wiking | aah ok | 15:20 |
| @HeikoS | and I use this battery hack | 15:20 |
| @HeikoS | only charge to 70% | 15:20 |
| @wiking | y? | 15:20 |
| @HeikoS | because on my old one the battery was bad quite quickly | 15:20 |
| @wiking | ah ok | 15:20 |
| @HeikoS | I only charge full on demand before flight | 15:20 |
| @wiking | i thought lithium likes to be more fully charged end | 15:21 |
| @HeikoS | i think they like 30-70 mist | 15:21 |
| @HeikoS | most | 15:21 |
| @HeikoS | there is a whole kernel module thingi for this stuff | 15:21 |
| @HeikoS | https://linrunner.de/en/tlp/docs/tlp-linux-advanced-power-management.html | 15:21 |
| @wiking | man not the cheapest laptop | 15:22 |
| @wiking | :) | 15:22 |
| @wiking | or soemthing is weird | 15:23 |
| @wiking | Lenovo ThinkPad X1 Carbon 6th Gen (14", Full HD, Intel Core i7-8550U, 16GB, SSD, 4G) | 15:24 |
| @HeikoS | so re the turing | 15:24 |
| @wiking | 2.3k uSD :D | 15:24 |
| @wiking | lolo | 15:24 |
| @HeikoS | Let me write this intro email | 15:24 |
| @HeikoS | yeah it is pricey | 15:24 |
| @wiking | ok gdoc it | 15:25 |
| @HeikoS | will do | 15:25 |
| @wiking | and lets take it from there | 15:25 |
| rcurtin | wiking: 89 :-\ | 15:26 |
| @wiking | rcurtin: lololo | 15:26 |
| @wiking | rcurtin: we are clearly doing something wronng :D | 15:26 |
| rcurtin | well, maybe not | 15:26 |
| rcurtin | were you applications good? | 15:27 |
| rcurtin | we had an onslaught of extremely low quality students this year | 15:27 |
| @wiking | haven't had time to look into it | 15:27 |
| rcurtin | like, people who are not able to operate a mailing list | 15:27 |
| rcurtin | people who check out the code then ask for help because they can't find where the code i | 15:27 |
| @wiking | hhehe | 15:27 |
| rcurtin | is | 15:27 |
| rcurtin | people who ask if they can write Python for the project (...for the C++ library...) | 15:27 |
| @wiking | ah ok | 15:27 |
| @wiking | so crazy stuff | 15:27 |
| rcurtin | yeah, I mean, it happens every year but it was worse than usual it seemed | 15:27 |
| @wiking | HeikoS: when do we release? | 15:28 |
| rcurtin | we did get some great students to apply for sure, just, lots of noise | 15:28 |
| @wiking | the list is getting thiner and thiner on project list | 15:28 |
| @HeikoS | wiking: just forward the call for proposal | 15:28 |
| @wiking | HeikoS: i would say i wanna merge the new serialization and the windows dll fix (two enormous prs) | 15:28 |
| @HeikoS | 29 april | 15:28 |
| @HeikoS | wiking: yes merge those | 15:29 |
| @HeikoS | and then we can remove parameter stuff | 15:29 |
| @wiking | i guess gf712 wanna add the grid part on? | 15:29 |
| @wiking | *no? | 15:29 |
| @HeikoS | gil is almost done with that as well | 15:29 |
| @HeikoS | yes defo | 15:29 |
| @wiking | ok | 15:29 |
| @HeikoS | because then a lot of stuff can go | 15:29 |
| @HeikoS | but there is some refactoring work for the gradient stuff needed after | 15:29 |
| @wiking | geektoni: how likely u r done like around easter? | 15:29 |
| @HeikoS | although I think gf712 will be able to do that really quiclk | 15:29 |
| @HeikoS | giovannis project is 3 mar, apr, may | 15:29 |
| @wiking | yeye i know | 15:30 |
| @wiking | but the question is how much of that thing could go in | 15:30 |
| @wiking | within a week or so | 15:30 |
| @wiking | or do we wanna actually have anything of that in? | 15:30 |
| @HeikoS | I think that can be next release | 15:30 |
| @wiking | mmm how much change of api is that? | 15:30 |
| @HeikoS | I mean what is done can be in | 15:30 |
| @HeikoS | not much | 15:30 |
| @HeikoS | mmh | 15:30 |
| @HeikoS | well | 15:30 |
| @wiking | coz if abi changes then we need to 8.x it | 15:30 |
| @wiking | :) | 15:30 |
| @HeikoS | yeah ok | 15:30 |
| geektoni | wiking: if the factory PR goes in like today I could add the observer stuff to some methods in the next days | 15:31 |
| @HeikoS | would be good to have the first part in | 15:31 |
| geektoni | but surely not much | 15:31 |
| @wiking | k | 15:31 |
| @HeikoS | it would be good to have that as it fixes the api more or less | 15:31 |
| @wiking | so maybe then we aim for 1st of may or somethinng | 15:31 |
| @HeikoS | and then integrating it further can be done in subrelease | 15:31 |
| @wiking | instead of easter | 15:31 |
| @HeikoS | hopefully before gsoc :D | 15:31 |
| @wiking | yeah as always | 15:31 |
| @wiking | :D | 15:31 |
| @HeikoS | oh btw | 15:31 |
| @HeikoS | how can i make a room? | 15:32 |
| @wiking | ok anyhow would be good to get it done (release) | 15:32 |
| @HeikoS | for geektoni and wiking ? | 15:32 |
| @wiking | HeikoS: mkdir room | 15:32 |
| @HeikoS | need to discuss something | 15:32 |
| @HeikoS | mkdir room | 15:32 |
| @wiking | :D | 15:32 |
| @wiking | i mean just /j #whateveruwant | 15:32 |
| @wiking | and invite us | 15:32 |
| @HeikoS | join #mkdirroom | 15:32 |
| @HeikoS | pls come | 15:32 |
| geektoni | mmh does it work from the web interface also? | 15:33 |
| @wiking | geektoni: should | 15:33 |
| @HeikoS | yes | 15:33 |
| @wiking | just /joinn | 15:33 |
| gf712 | wiking: not sure yet if there will be errors with machine eval | 15:42 |
| gf712 | but I am having issues locally | 15:43 |
| gf712 | when I removed refcount from some | 15:43 |
| gf712 | so lets see | 15:43 |
| @wiking | what sorts of issues? | 15:43 |
| gf712 | labels is deleted is machine eval | 15:43 |
| gf712 | I am guessing ref count issue | 15:43 |
| @wiking | miju | 15:43 |
| @wiking | ok i'll look into it | 15:44 |
| @wiking | i was doing almost like a regexp | 15:44 |
| @wiking | for removing those things :) | 15:44 |
| gf712 | haha | 15:44 |
| gf712 | can you not do some fancy lib tooling? | 15:44 |
| gf712 | wiking: I might be wrong | 15:44 |
| @wiking | you can always do that | 15:44 |
| @wiking | :))) | 15:44 |
| gf712 | well see | 15:44 |
| @wiking | so far things are fine on osx | 15:44 |
| @wiking | but that's only c++ stuff | 15:45 |
| @wiking | so let's see swig soonish | 15:45 |
| gf712 | but osx doesn't test meta right? | 15:45 |
| @HeikoS | "Multi-language distributed high-performance model selection with Shogun.ML" | 15:45 |
| @wiking | gf712: only c++ afaik | 15:45 |
| @wiking | or maybe not even that? | 15:45 |
| gf712 | only unitest there | 15:46 |
| gf712 | I think | 15:46 |
| @wiking | gf712: damn shame :) | 15:46 |
| gf712 | and gcc does meta | 15:46 |
| gf712 | well linux | 15:46 |
| gf712 | wiking: seems to be all good | 15:47 |
| gf712 | I'll look into it locally | 15:48 |
| gf712 | not sure what happened | 15:48 |
| @wiking | gcc is still compiling | 15:48 |
| @HeikoS | gf712, wiking btw we should also do some kind of wrap up for the coreml/openml project | 15:50 |
| @HeikoS | with a page that contains the abstract, a demo | 15:50 |
| @wiking | HeikoS: yes | 15:50 |
| @HeikoS | to send around to peeps | 15:50 |
| @wiking | HeikoS: working on the react app | 15:50 |
| @wiking | :)) | 15:50 |
| @wiking | that's the ultimate demo | 15:50 |
| @HeikoS | not so technical, but rather temo | 15:50 |
| @HeikoS | ah cool | 15:50 |
| @wiking | you have a web browser | 15:51 |
| @wiking | you click shit around | 15:51 |
| @HeikoS | well it needs to be accessilble to academics | 15:51 |
| @wiking | u get a model | 15:51 |
| @wiking | and then u export | 15:51 |
| @wiking | to coreml or openml | 15:51 |
| @HeikoS | that is good | 15:51 |
| @wiking | HeikoS: https://js.shogun.ml good? :) | 15:51 |
| @wiking | u put that into the browser and that's it | 15:51 |
| @HeikoS | ++ | 15:51 |
| @HeikoS | ok let me know once that goes | 15:51 |
| @HeikoS | I will think about text etc | 15:51 |
| @wiking | gf712: ok gcc passed | 15:51 |
| @wiking | gf712: lets see the swigs | 15:52 |
| @wiking | gf712: i'll merge this regardless what's in valgrind :P | 15:52 |
| @wiking | if ci passes | 15:52 |
| gf712 | yea, it will be def better | 15:52 |
| @wiking | coz this is more correct than the current develop approach | 15:52 |
| gf712 | it shouldn't leak as much | 15:52 |
| @wiking | as in case of factory | 15:52 |
| @wiking | semantically sg-refing meanign that factory takes ref to that shit | 15:52 |
| @wiking | right? | 15:52 |
| @wiking | which indeed is wrong | 15:53 |
| @wiking | factory should not hold ref to the created object | 15:53 |
| gf712 | yea, I guess so! | 15:53 |
| @wiking | the consumer shoudl decide where and who holds the ref | 15:53 |
| @wiking | dont gess | 15:53 |
| gf712 | true | 15:53 |
| @wiking | *guess | 15:53 |
| @wiking | :D | 15:53 |
| @wiking | DO | 15:53 |
| @wiking | ok gotta run again for 20 mins | 15:53 |
| @wiking | ttyl | 15:53 |
| gf712 | aight | 15:53 |
| gf712 | ttyl | 15:53 |
| @HeikoS | see you later | 15:55 |
| @HeikoS | also need to go | 15:55 |
| -!- wiking [~wiking@huwico/staff/wiking] has quit [Read error: Connection reset by peer] | 15:55 | |
| @HeikoS | will check in back later as well | 15:55 |
| @HeikoS | gf712: how is the grid stuff going? | 15:55 |
| gf712 | HeikoS: just fixing some minor stuff | 15:56 |
| gf712 | api is the same/similar to the old one | 15:56 |
| gf712 | but need to rethink what the base class should be | 15:56 |
| gf712 | right now I use multiple inheritance | 15:56 |
| @HeikoS | mixins maybe? | 15:57 |
| @HeikoS | that works with swig | 15:57 |
| gf712 | yea, I think so | 15:57 |
| @HeikoS | compile time might blow up though | 15:57 |
| gf712 | because there are two seperate abstractions | 15:57 |
| @HeikoS | but there are not that many classes which use these abstractions or? | 15:57 |
| gf712 | the parameter combination and the execution engine | 15:57 |
| gf712 | no | 15:57 |
| gf712 | but the parameter combination would be cool to use | 15:58 |
| gf712 | for serialising to openml | 15:58 |
| gf712 | for example | 15:58 |
| @HeikoS | yes | 15:59 |
| gf712 | basically I need to have a container that is similar to sgobject | 15:59 |
| @HeikoS | it would be great to re-use that stuff | 15:59 |
| @HeikoS | https://github.com/shogun-toolbox/shogun/pull/4598 | 15:59 |
| gf712 | but not quite | 15:59 |
| @HeikoS | this is the pr right? | 15:59 |
| gf712 | yes | 15:59 |
| @HeikoS | ok | 15:59 |
| @HeikoS | i will read a bit | 15:59 |
| gf712 | ParameterNode represent a combination | 15:59 |
| gf712 | and then GridParameters represents the space | 16:00 |
| gf712 | from which you can generate ParameterNode | 16:00 |
| @HeikoS | ah ok cool | 16:01 |
| @HeikoS | so then you can have different way to generate params | 16:01 |
| gf712 | yea, so GridParameters has vectors | 16:01 |
| @HeikoS | some strategies will require some additional information | 16:01 |
| gf712 | instead of scalar values | 16:01 |
| @HeikoS | like grid-serach requires a search space bounds or values | 16:01 |
| @HeikoS | gradient search requires a diff-able function | 16:01 |
| @HeikoS | etc | 16:01 |
| gf712 | yup, which you can just add | 16:01 |
| gf712 | I had to add iterator pointers for example | 16:01 |
| gf712 | which don't make sense in ParameterNode | 16:02 |
| @HeikoS | i see | 16:02 |
| gf712 | but are required in GridParameters | 16:02 |
| @HeikoS | yes | 16:02 |
| @HeikoS | that was awkward in the old design | 16:02 |
| gf712 | so then the attach methods are virtual | 16:02 |
| @HeikoS | i need to read to understand | 16:03 |
| gf712 | so that the use can't add a scalar to GridParametrs | 16:03 |
| gf712 | OK! | 16:03 |
| @HeikoS | I am just pointing on the gradient stuff all the time so we can put that under this new hood as well | 16:03 |
| @HeikoS | the whole GP optimization framework | 16:03 |
| gf712 | yea, it should be fine | 16:04 |
| gf712 | at least I think so | 16:04 |
| gf712 | I would have to work on that to understand it better | 16:04 |
| gf712 | but the base class is not too restrictive | 16:05 |
| @HeikoS | the main change is that | 16:05 |
| @HeikoS | a) there are no provided boundaries or values, but rather just parameter names and maybe starting values | 16:05 |
| @HeikoS | b) it doesnt use cross-validation to check but something else, the marginal likelihood which is implemented somewhere in the gp framework, needs to be bridged api wise | 16:06 |
| @HeikoS | c) there is no loop over different parameter combinations that are know a-priori, but rather only once you have evaluated one parameter combination you can generate a new one | 16:06 |
| gf712 | a) that is fine | 16:07 |
| gf712 | b) right now that is something that the engine decides so that can be changed easily | 16:07 |
| gf712 | b) the train method is pure virtual so has to be rewritten for each type of model selection | 16:07 |
| gf712 | c) I guess that kinda goes with b) | 16:08 |
| @HeikoS | gf712: I have one more thing. Check out the IterativeMachine mixin. This implements a stoppable training method. It would be very cool if the new modelselection was also stoppable, then then the yet-best-parameters would be extractable. But can discuss that later | 16:08 |
| @HeikoS | I gotta run now. Talk in a bit | 16:08 |
| gf712 | OK! should be doable | 16:09 |
| gf712 | ttyl | 16:09 |
| -!- HeikoS [~heiko@82-132-215-5.dab.02.net] has quit [Remote host closed the connection] | 16:10 | |
| -!- gf712 [90520840@gateway/web/freenode/ip.144.82.8.64] has quit [Ping timeout: 256 seconds] | 16:36 | |
| -!- geektoni [c1cdd24a@gateway/web/freenode/ip.193.205.210.74] has quit [Quit: Page closed] | 16:53 | |
| -!- wiking [~wiking@178.197.237.54] has joined #shogun | 16:59 | |
| -!- wiking [~wiking@178.197.237.54] has quit [Changing host] | 16:59 | |
| -!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 16:59 | |
| -!- mode/#shogun [+o wiking] by ChanServ | 16:59 | |
| -!- gf712 [1f0ef8ff@gateway/web/freenode/ip.31.14.248.255] has joined #shogun | 17:21 | |
| gf712 | wiking: why doesn't [skip ci] work? | 17:23 |
| @wiking | ask m$ | 17:23 |
| @wiking | non idea | 17:23 |
| gf712 | hmm | 17:23 |
| @wiking | and my n is still stuck | 17:23 |
| @wiking | :D | 17:23 |
| gf712 | your what? | 17:24 |
| @wiking | key n | 17:24 |
| @wiking | on my laptop | 17:24 |
| gf712 | oh right haha | 17:24 |
| @wiking | this new macbook keyboard is a joke | 17:25 |
| gf712 | do yo have pre? | 17:25 |
| gf712 | pro | 17:25 |
| @wiking | ye | 17:25 |
| gf712 | hmm, can't complain with mine | 17:26 |
| gf712 | at least hardware | 17:26 |
| @wiking | how old? | 17:26 |
| gf712 | the Touch Bar is pretty useless though | 17:26 |
| gf712 | 5 months | 17:26 |
| @wiking | yeah that's for show | 17:26 |
| @wiking | ok | 17:26 |
| @wiking | so wait another 8-10 months | 17:26 |
| @wiking | and you'll see what i | 17:26 |
| @wiking | i'm talking about | 17:27 |
| gf712 | its the ATI's so all good :D | 17:27 |
| @wiking | :>>>>> | 17:27 |
| @wiking | yeah i mean this is not mine either | 17:27 |
| @wiking | but in order to get this resolved | 17:27 |
| @wiking | i would need to bring it to apple and have them fix it | 17:27 |
| @wiking | that takes 5 days | 17:27 |
| @wiking | or so | 17:27 |
| @wiking | which makes me look at the wall | 17:27 |
| @wiking | :) | 17:27 |
| gf712 | do you not have a desktop? | 17:27 |
| @wiking | nojp | 17:28 |
| @wiking | i only have this and a 40 inch 4k monitor :D | 17:28 |
| gf712 | haha | 17:28 |
| gf712 | get the ATI to buy you one | 17:28 |
| @wiking | lol | 17:28 |
| gf712 | not sure they'll ship it there though :p | 17:28 |
| @wiking | yeah breeeexit0 | 17:29 |
| gf712 | what? | 17:30 |
| @wiking | well | 17:31 |
| @wiking | first of all | 17:31 |
| @wiking | .ch is not eu | 17:31 |
| @wiking | and neither uk now i guess | 17:31 |
| @wiking | so shipping a computer would be like crazy taxes | 17:31 |
| @wiking | :D | 17:31 |
| gf712 | oh to ship it | 17:31 |
| gf712 | yea but you can pick it up and then sail back with it | 17:31 |
| @wiking | ahhahahah | 17:32 |
| @wiking | gf712: ok | 17:35 |
| @wiking | i have a q | 17:35 |
| @wiking | how do you think i should newobject these ones | 17:36 |
| @wiking | %template(features) shogun::features<float64_t>; | 17:36 |
| @wiking | :D | 17:36 |
| @wiking | or probably %newobject shogun::features<float64_t>; | 17:37 |
| @wiking | let's see | 17:37 |
| gf712 | %newobject shogun::features<float64_t>; | 17:40 |
| @wiking | yeah trying | 17:40 |
| gf712 | hmm | 17:40 |
| @wiking | not really hekping :)) | 17:41 |
| gf712 | actually | 17:41 |
| gf712 | why? | 17:41 |
| gf712 | you already have %newobject shogun::features | 17:41 |
| gf712 | no> | 17:41 |
| gf712 | ? | 17:41 |
| @wiking | yeah but for different function | 17:41 |
| @wiking | it's not the same shit | 17:41 |
| @wiking | i have | 17:41 |
| @wiking | %newobject shogun::features(CFile*, EPrimitiveType primitive_type); | 17:41 |
| @wiking | %newobject shogun::features(const std::string& name); | 17:41 |
| @wiking | that does not cover feature<>lalal | 17:42 |
| @wiking | ok be right back | 17:42 |
| -!- wiking [~wiking@huwico/staff/wiking] has quit [Remote host closed the connection] | 17:42 | |
| -!- gf712 [1f0ef8ff@gateway/web/freenode/ip.31.14.248.255] has quit [Ping timeout: 256 seconds] | 18:29 | |
| -!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 21:30 | |
| -!- mode/#shogun [+o wiking] by ChanServ | 21:30 | |
| --- Log closed Thu Apr 11 00:00:31 2019 | ||
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!