--- Log opened Mon Apr 29 00:00:53 2019 | ||
-!- wiking [~wiking@c-185-45-237-122.customer.ggaweb.ch] has quit [Remote host closed the connection] | 01:48 | |
-!- wiking [~wiking@c-185-45-237-122.customer.ggaweb.ch] has joined #shogun | 01:51 | |
-!- wiking [~wiking@c-185-45-237-122.customer.ggaweb.ch] has quit [Remote host closed the connection] | 06:51 | |
-!- wiking [~wiking@c-185-45-237-122.customer.ggaweb.ch] has joined #shogun | 06:55 | |
-!- wiking [~wiking@c-185-45-237-122.customer.ggaweb.ch] has quit [Ping timeout: 245 seconds] | 07:00 | |
-!- wiking [~wiking@c-185-45-237-122.customer.ggaweb.ch] has joined #shogun | 07:00 | |
-!- wiking [~wiking@c-185-45-237-122.customer.ggaweb.ch] has quit [Remote host closed the connection] | 07:03 | |
-!- wiking [~wiking@c-185-45-237-122.customer.ggaweb.ch] has joined #shogun | 07:03 | |
-!- wiking [~wiking@c-185-45-237-122.customer.ggaweb.ch] has quit [Ping timeout: 246 seconds] | 07:08 | |
-!- wiking [~wiking@c-185-45-237-122.customer.ggaweb.ch] has joined #shogun | 07:45 | |
-!- wiking [~wiking@c-185-45-237-122.customer.ggaweb.ch] has quit [Remote host closed the connection] | 08:22 | |
-!- wiking [~wiking@c-185-45-237-122.customer.ggaweb.ch] has joined #shogun | 08:24 | |
-!- wiking [~wiking@c-185-45-237-122.customer.ggaweb.ch] has quit [Ping timeout: 246 seconds] | 08:30 | |
-!- gf712 [c13cdcfd@gateway/web/freenode/ip.193.60.220.253] has joined #shogun | 09:51 | |
-!- sighingnow [~sighingno@2001:da8:203:81:381f:7109:7db8:1e97] has joined #shogun | 11:08 | |
-!- sighingnow [~sighingno@2001:da8:203:81:381f:7109:7db8:1e97] has left #shogun [] | 11:08 | |
-!- HeikoS [~heiko@52.pool85-48-187.static.orange.es] has joined #shogun | 14:27 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 14:27 | |
@HeikoS | gf712: yo! | 14:27 |
---|---|---|
gf712 | HeikoS: hey | 15:08 |
@HeikoS | hey | 15:09 |
gf712 | viktor says he'll be there in a sec | 15:09 |
-!- wiking_ [~wiking@193.60.220.253] has joined #shogun | 15:09 | |
@HeikoS | ah here we go :) | 15:09 |
@HeikoS | so I think we are in pretty good shape proposal wise | 15:09 |
wiking_ | ok | 15:09 |
wiking_ | here | 15:09 |
@HeikoS | wiking_: hey | 15:09 |
gf712 | cool, need to reread a bit | 15:09 |
wiking_ | ok so whatsapp | 15:10 |
wiking_ | with natalie | 15:10 |
wiking_ | i told her that we might talk | 15:10 |
wiking_ | imo that 2 weeks is too little | 15:10 |
@HeikoS | yeah sure | 15:10 |
@HeikoS | 4 weeks is better | 15:10 |
wiking_ | once it's settled with the numbers | 15:10 |
@HeikoS | I just wanted to discuss with you: whose name to put on her proposal | 15:10 |
@HeikoS | and the numbers | 15:10 |
wiking_ | look | 15:10 |
wiking_ | since atm | 15:10 |
wiking_ | i have the least amount of nunmbers | 15:10 |
wiking_ | from the 3 of us | 15:10 |
wiking_ | in the other proposal | 15:10 |
wiking_ | i dont care | 15:11 |
wiking_ | but from ATI perspecitve | 15:11 |
wiking_ | like if we add Gil then they'll be ?? | 15:11 |
wiking_ | i guess | 15:11 |
wiking_ | see what i'm heading too? | 15:11 |
@HeikoS | yes | 15:11 |
wiking_ | *to | 15:11 |
@HeikoS | even though Gil should be involved in meetings/workshops | 15:11 |
@HeikoS | wiking_: no I dont see yet | 15:11 |
wiking_ | ok so then continuing | 15:11 |
wiking_ | if we add Gil who's currently 75% | 15:12 |
wiking_ | then 4 weeks | 15:12 |
wiking_ | FTE | 15:12 |
wiking_ | = 20 weeks with 20% | 15:12 |
wiking_ | right? | 15:12 |
wiking_ | or 16 weeks 25% | 15:12 |
wiking_ | we cannot go higher | 15:12 |
wiking_ | as in the other application there's already 75% allocated on him | 15:12 |
wiking_ | so if both applications gets accepted | 15:13 |
wiking_ | then there's a race condition on the resource | 15:13 |
@HeikoS | but this is one -off | 15:13 |
@HeikoS | 4 weeks | 15:13 |
wiking_ | yes | 15:13 |
wiking_ | but that is still some amount of FTE | 15:13 |
@HeikoS | okok | 15:13 |
wiking_ | as part of a project | 15:13 |
@HeikoS | so what do you suggest? | 15:13 |
wiking_ | or package | 15:13 |
wiking_ | so to avoid these sort of concerns | 15:13 |
wiking_ | (resource allocation and overallocation) | 15:14 |
wiking_ | i would say add me | 15:14 |
@HeikoS | cool with me | 15:14 |
wiking_ | if somebody needs the$ | 15:14 |
@HeikoS | so let's do that then | 15:14 |
wiking_ | i'm happy to give transfer it to that person | 15:14 |
@HeikoS | no I am happy for you to be | 15:14 |
wiking_ | as funding wise i dont care | 15:14 |
@HeikoS | think this is more the names on the proposals, that is the more important thing | 15:15 |
@HeikoS | generates momentum | 15:15 |
wiking_ | k lets see what natalie says | 15:15 |
@HeikoS | ok I have an email ready | 15:15 |
@HeikoS | oh no actually I sent it | 15:15 |
@HeikoS | :) | 15:15 |
@HeikoS | but no numbers yes | 15:15 |
@HeikoS | wiking_: Ill send her 4 weeks of your time then? | 15:16 |
wiking_ | lets wait | 15:17 |
wiking_ | for Natalie | 15:17 |
@HeikoS | ok | 15:17 |
@HeikoS | Ill leave it to you then to send the $$$ | 15:17 |
wiking_ | yep i just wanna see | 15:18 |
wiking_ | that she is ok with more than 2 weeks | 15:18 |
wiking_ | as that can do it | 15:18 |
wiking_ | but it's very very optimistic | 15:18 |
wiking_ | HeikoS: we've just went through the todo with gf712 till end of may | 15:19 |
wiking_ | i need to get deadbeef merged | 15:19 |
wiking_ | as we need tagging based xml serialization | 15:19 |
wiking_ | for openml | 15:19 |
wiking_ | he is working on ARFF format | 15:19 |
@HeikoS | great | 15:19 |
wiking_ | i'll come end of may again | 15:20 |
@HeikoS | wiking_: I wanted to fix some bugs | 15:20 |
wiking_ | but i. mean | 15:20 |
wiking_ | currently | 15:20 |
@HeikoS | as there is some issues with labels | 15:20 |
wiking_ | XVAl | 15:20 |
wiking_ | is LEAKYING like nothing | 15:20 |
wiking_ | :DDD | 15:20 |
@HeikoS | ah yes | 15:20 |
wiking_ | so that's another one | 15:20 |
wiking_ | but lisitsyn is on las palmas | 15:20 |
wiking_ | and i dont know what to do with | 15:20 |
wiking_ | make_any(shared_ptr<>) | 15:20 |
wiking_ | ideae? | 15:21 |
@HeikoS | ah yes | 15:21 |
wiking_ | *ideas? | 15:21 |
@HeikoS | not really | 15:21 |
@HeikoS | but what I know | 15:21 |
@HeikoS | is that a lot of stuff inside any.h | 15:21 |
@HeikoS | is based on raw pointers | 15:21 |
@HeikoS | as we needed to make it compatible with old arrays | 15:21 |
@HeikoS | ah all that fuck | 15:21 |
@HeikoS | btw | 15:21 |
@HeikoS | saw the gsod stuff? | 15:22 |
wiking_ | yes | 15:22 |
@HeikoS | I like that quite a bit as well | 15:22 |
@HeikoS | but maybe it is too much | 15:22 |
@HeikoS | not sure | 15:22 |
@HeikoS | wiking_, gf712 I will submit the proposal later today then | 15:23 |
@HeikoS | after we heard back from natalie | 15:23 |
@HeikoS | and then I think it is done | 15:23 |
@HeikoS | I copy it into the form, shorten (max words etc), and do a final polish | 15:23 |
@HeikoS | but I think it looks good | 15:23 |
wiking_ | w8 | 15:24 |
@HeikoS | and it is great to have 3 collaborators on board as well ... so lets see | 15:24 |
wiking_ | ah yeah nothing | 15:24 |
wiking_ | yeah we can get natalie on board there | 15:24 |
wiking_ | imo | 15:24 |
@HeikoS | yes | 15:24 |
@HeikoS | i added her as 0.02 FTE | 15:24 |
@HeikoS | 3 hrs/ month | 15:24 |
@HeikoS | meetings | 15:24 |
@HeikoS | she can tell us what is needed | 15:24 |
@HeikoS | and this way we get some idea of user perspective | 15:24 |
@HeikoS | and also have another colab | 15:24 |
@HeikoS | win-win | 15:24 |
@HeikoS | gf712: wiking_ alright I'll leave you two to it. Will keep an eye open on my emails | 15:29 |
gf712 | HeikoS: ok, I'll be finising up the openml stuff | 15:30 |
gf712 | we concluded that it might be best to write our own stuff with shogun | 15:30 |
@HeikoS | gf712: cool, can I help with anything regarding blog-post/ write-up | 15:30 |
gf712 | so that we can expose openml via swig | 15:30 |
@HeikoS | gf712: can you explain? | 15:30 |
gf712 | so right now we use openml-python | 15:31 |
gf712 | but that means it is only accessible via the openml library for python | 15:31 |
gf712 | but if you use shogun with another language it won't work | 15:31 |
gf712 | so we are our own thing to convert shogun to the required stuff for openml | 15:31 |
gf712 | we have all the things required | 15:32 |
@HeikoS | ah I see | 15:32 |
@HeikoS | yeah that is nice | 15:32 |
gf712 | just need to write the ARFF parser | 15:32 |
@HeikoS | then one can export to openml from Java bindings | 15:32 |
gf712 | which is pretty straight forward | 15:32 |
gf712 | yes | 15:32 |
@HeikoS | great idea! | 15:32 |
@HeikoS | like it | 15:32 |
gf712 | I was going to do it for the next project | 15:32 |
@HeikoS | hehe | 15:32 |
gf712 | but then the thing openml happened | 15:32 |
gf712 | where they changed a bunch of stuff | 15:32 |
gf712 | so since I have to refactor | 15:33 |
gf712 | I might as well do it in our end | 15:33 |
@HeikoS | I see | 15:33 |
gf712 | and then no need to worry anymore | 15:33 |
gf712 | Viktor will sort out the rest api stuff | 15:33 |
gf712 | so just need to serialise to xml and arff | 15:33 |
gf712 | and then can send that to the server! | 15:34 |
@HeikoS | sweet | 15:34 |
gf712 | and the people who wanted to use R and shogun from that proposal can use openml | 15:34 |
@HeikoS | wiking_: natalie replied (to me only), can you chat now, or at 13:45, or at 14:00? | 15:36 |
lisitsyn | wiking_: hey | 15:39 |
wiking_ | ho | 15:40 |
lisitsyn | is it better to chat here? :) | 15:40 |
wiking_ | HeikoS: anytime | 15:40 |
wiking_ | just tell me when | 15:40 |
wiking_ | lisitsyn: yes | 15:40 |
@HeikoS | wiking_: ok I suggested 14:00, will ping you | 15:40 |
wiking_ | lisitsyn: so basically with the current way | 15:40 |
wiking_ | if i pass &(shared_ptr<T> ) to param framework | 15:40 |
wiking_ | then in case of clone | 15:40 |
wiking_ | the copy-ctor is being called | 15:41 |
wiking_ | and that's not good | 15:41 |
wiking_ | :) | 15:41 |
lisitsyn | why? | 15:41 |
wiking_ | coz then underlying pointer is not cloned | 15:41 |
lisitsyn | ahh ok | 15:41 |
wiking_ | its just another shared_ptr | 15:41 |
@HeikoS | I think that is easy to add actually | 15:41 |
wiking_ | over the same ptr | 15:41 |
lisitsyn | easy then | 15:41 |
wiking_ | ok i'm listening | 15:41 |
wiking_ | :D | 15:42 |
lisitsyn | let me point | 15:42 |
lisitsyn | wiking_: see how that's done for ArrayReference https://github.com/shogun-toolbox/shogun/blob/develop/src/shogun/lib/any.h#L321 | 15:42 |
wiking_ | yep | 15:43 |
wiking_ | template <class T> | 15:47 |
wiking_ | inline auto clone(void** storage, const std::shared_ptr<T>& value) | 15:47 |
wiking_ | { | 15:47 |
wiking_ | auto existing = mutable_value_of<std::shared_ptr<T>>(storage); | 15:47 |
wiking_ | return std::shared_ptr<T>(clone(existing)); | 15:47 |
wiking_ | } | 15:47 |
wiking_ | ? | 15:47 |
lisitsyn | yes | 15:48 |
wiking_ | and then the clone of existing will create the clone of sgo? | 15:48 |
lisitsyn | I guess so | 15:48 |
wiking_ | kk | 15:48 |
wiking_ | btw | 15:48 |
wiking_ | actually | 15:48 |
wiking_ | currently SGO::clone in that branch | 15:48 |
wiking_ | returns a shared_ptr :P | 15:48 |
wiking_ | u mind? | 15:48 |
wiking_ | so not SGO* anymore | 15:49 |
wiking_ | because of the ref story | 15:49 |
wiking_ | but semantically its a bit shitty | 15:49 |
lisitsyn | while it works I am fine :) | 15:50 |
wiking_ | ok lemme try to add this to any | 15:50 |
wiking_ | and see how it blows up | 15:50 |
wiking_ | :D | 15:50 |
wiking_ | lisitsyn: https://pastebin.com/Pb2Jn6bc | 15:57 |
wiking_ | i think i know why | 15:58 |
lisitsyn | wiking_: there is no clone so you should call it explicitly | 16:14 |
lisitsyn | I mean the clone in any is like a function with a side-effect | 16:14 |
lisitsyn | I guess you just need ->clone() | 16:14 |
wiking_ | yes | 16:16 |
wiking_ | just doing that | 16:16 |
wiking_ | :) | 16:16 |
wiking_ | compiling | 16:16 |
wiking_ | ... | 16:16 |
wiking_ | :D | 16:16 |
wiking_ | lisitsyn: we have a tiny problem | 16:41 |
wiking_ | say we have LinearMachine | 16:41 |
lisitsyn | yes | 16:42 |
wiking_ | that has std::shared_ptr<CDotFeatures> features; | 16:42 |
wiking_ | but we wanna register it as std::shared_ptr<CFeatures> | 16:42 |
wiking_ | because of put/get | 16:42 |
wiking_ | right? | 16:42 |
wiking_ | but because it's a linear machine | 16:42 |
wiking_ | we wanna have a CDotFeatures | 16:42 |
wiking_ | so if i do this right | 16:43 |
wiking_ | watch_param("features", &(std::static_pointer_cast<CFeatures>(features))); | 16:43 |
wiking_ | then of course the compiler is gonna bitch | 16:43 |
wiking_ | /Users/wiking/shogun/src/shogun/machine/LinearMachine.cpp:40:25: error: taking the address of a temporary object of type 'std::shared_ptr<CFeatures>' [-Waddress-of-temporary] | 16:43 |
wiking_ | but if i dont cast | 16:43 |
wiking_ | then its not a base param | 16:44 |
wiking_ | (explicit cast doesn't work :P | 16:44 |
wiking_ | since its shared_ptr right :) | 16:44 |
wiking_ | buuut ok lets forget this | 16:48 |
wiking_ | basically there's another problem | 16:48 |
wiking_ | auto existing = mutable_value_of<std::shared_ptr<T>>(storage); | 16:48 |
wiking_ | storage is non null | 16:48 |
wiking_ | but the casting above returns a null ptr | 16:48 |
wiking_ | (lldb) p existing | 16:49 |
wiking_ | (std::__1::shared_ptr<shogun::CLabels>) $0 = nullptr { | 16:49 |
wiking_ | __ptr_ = 0x0000000000000000 | 16:49 |
wiking_ | } | 16:49 |
wiking_ | (lldb) p storage | 16:49 |
wiking_ | (void **) $1 = 0x00007ffeefbfec48 | 16:49 |
wiking_ | lisitsyn: ^ | 16:49 |
lisitsyn | uh | 16:50 |
wiking_ | ah no | 16:50 |
wiking_ | this is not good :) | 16:50 |
wiking_ | something is wrong with this... as value is not being used at all | 16:50 |
wiking_ | template <class T> | 16:50 |
wiking_ | inline auto clone(void** storage, const std::shared_ptr<T>& value) | 16:50 |
wiking_ | { | 16:50 |
wiking_ | auto existing = mutable_value_of<std::shared_ptr<T>>(storage); | 16:50 |
wiking_ | return existing->clone(); | 16:51 |
wiking_ | } | 16:51 |
wiking_ | what storage should hold in this case? | 16:51 |
wiking_ | or that is the "new" storage | 16:52 |
lisitsyn | I think it is fine for storage to hold an empty thing | 16:52 |
lisitsyn | yeah | 16:52 |
wiking_ | aaah ok | 16:52 |
wiking_ | so storage.reset(value->clone()) | 16:52 |
wiking_ | no? | 16:52 |
lisitsyn | yes I think so | 16:52 |
lisitsyn | uhmm yes | 16:52 |
lisitsyn | should work | 16:52 |
lisitsyn | as for the first problem it is getting tricky | 16:53 |
-!- HeikoS1 [~heiko@200.pool85-48-187.static.orange.es] has joined #shogun | 17:13 | |
-!- HeikoS [~heiko@52.pool85-48-187.static.orange.es] has quit [Ping timeout: 246 seconds] | 17:14 | |
-!- HeikoS1 [~heiko@200.pool85-48-187.static.orange.es] has quit [Quit: Leaving.] | 17:18 | |
-!- gf712 [c13cdcfd@gateway/web/freenode/ip.193.60.220.253] has quit [Ping timeout: 256 seconds] | 18:24 | |
wiking_ | lisitsyn: can we have rvalue for make_any_ref and then use other policy than non-owninng? | 18:46 |
wiking_ | :) | 18:46 |
wiking_ | lisitsyn: that case i think this casting should be solved | 18:46 |
wiking_ | as then we would take the temp object as rvalue (as it is rvalue) | 18:46 |
wiking_ | and anyways the underlying thingy is shared between the original shared_ptr and the casted one | 18:46 |
wiking_ | see what i mean? | 18:47 |
wiking_ | so the other case (no casting) clone works with std::shared_ptr now \o/ | 18:50 |
-!- wiking_ [~wiking@193.60.220.253] has quit [Remote host closed the connection] | 18:56 | |
-!- HeikoS [~heiko@228.pool85-48-187.static.orange.es] has joined #shogun | 19:03 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 19:03 | |
-!- HeikoS [~heiko@228.pool85-48-187.static.orange.es] has quit [Ping timeout: 250 seconds] | 19:29 | |
-!- HeikoS [~heiko@58.pool85-48-187.static.orange.es] has joined #shogun | 19:30 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 19:30 | |
-!- HeikoS [~heiko@58.pool85-48-187.static.orange.es] has quit [Ping timeout: 268 seconds] | 19:40 | |
-!- HeikoS [~heiko@153.pool85-48-187.static.orange.es] has joined #shogun | 19:42 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 19:42 | |
-!- HeikoS [~heiko@153.pool85-48-187.static.orange.es] has quit [Client Quit] | 19:44 | |
-!- rsdel2007 [75dc70b6@gateway/web/freenode/ip.117.220.112.182] has joined #shogun | 22:29 | |
-!- rsdel2007 [75dc70b6@gateway/web/freenode/ip.117.220.112.182] has quit [Client Quit] | 22:30 | |
--- Log closed Tue Apr 30 00:00:55 2019 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!