04:24 < PrinceGuptaGitte> since training networks on datasets like ImageNet is practically impossible (unless I have a powerful work station), is there a way to transfer pre trained weights to MLPack models?

05:41 < kartikdutt18Gitt> @prince776 take a look at https://github.com/sreenikSS/mlpack-Tensorflow-Translator . Since not every layer is supported we can't really exchange weights between all layers.

05:43 < kartikdutt18Gitt> Also if you set up mlpack remotely with a system with CUDA you can use NVBLAS.

08:03 < jenkins-mlpack2> Project docker mlpack nightly build build #631: STILL FAILING in 2 hr 49 min: http://ci.mlpack.org/job/docker%20mlpack%20nightly%20build/631/

12:31 < rcurtin> well... that's unfortunate

12:32 < rcurtin> I now have a meeting with my company's CEO exactly at the mlpack video meet-up time

12:32 < rcurtin> it should only be half an hour though, so I'll join, but I'll probably be late (unless our meeting goes over...)

12:32 < rcurtin> I think the CEO is the one person I shouldn't reschedule with :)

12:33 < zoq> rcurtin: I don't mind to postpone, might not work for me as well.

12:33 < rcurtin> do you mean, e.g., postpone for like half an hour, or for like a week?

12:34 < zoq> a week

12:34 < zoq> But if it works for some people, fine for me.

12:57 < rcurtin> hmm, I don't think that the two of us need to be there every time... how about we just do this week at the scheduled time, and then do two weeks from now?

12:57 < rcurtin> it's totally casual so I don't think it's a big deal either way

12:58 < zoq> fine for me, I might join the meeting, but I don't know for sure

13:10 < rcurtin> likewise, I'll send an email to point that out

13:11 < zoq> Sounds good.

13:15 < AbishaiEbenezerG> for how long could the video chat be? i think the timing of the meetup is around 11:30 IST where i stay...

13:16 < sreenik[m]> freenode_gitter_abishaiema[m]: It is generally an hour long, though you can join or leave at any time you wish

13:17 < AbishaiEbenezerG> cool

13:37 < PrinceGuptaGitte> Finally I'll attend it this time

14:06 < M_slack_9> Would you please give the details of the video chat?

14:06 < M_slack_9> I am interested to join.

14:58 < rcurtin> M_slack_9: details on the website: https://www.mlpack.org/community.html#real-time-chat

14:58 < rcurtin> just got those deployed :)

15:00 < mohona[m]> Thank you!

15:12 < Param-29Gitter[m> I was working on decision trees but cannot understand what I see in results. First I tried to speed up classify function. Its execution time did decrease but it takes more time to train (the model) with increase in threads.

15:12 < Param-29Gitter[m> Can someone help me understand why this is happening?

15:13 < Param-29Gitter[m> training time (1T - 59s, 4T- 75s) testing time (1T - 0.11s, 4T - 0.051s)

15:53 < saksham189Gitter> Hey @zoq I wanted your opinion regarding the adaptive pooling mean and max layers PR ( https://github.com/mlpack/mlpack/pull/2195 ). Do you think we should implement the layer as a wrapper over the original pooling layer since most of the code is exactly the same once the stride and kernel parameters have been calculated?

16:05 < PrinceGuptaGitte> Hi @kartikdutt18 thanks for providing the mlpack-tensorflow-translator's source. I'm able to get a general idea of how it's working. Do you have some code in which a keras model saved as onnx is being loaded into mlpack.

16:30 < kartikdutt18Gitt> At this point I don't have a code for that right now. Maybe Sreenik might be able to help you with that.

16:31 < sreenik[m]> freenode_gitter_prince776[m]: Hello, the convert_model() function in https://github.com/sreenikSS/mlpack-Tensorflow-Translator/blob/master/src/onnx_to_mlpack.hpp is what you are looking for

16:32 < sreenik[m]> Let me know if it does not produce expected results for your model, I am not 100% sure that it is not buggy

16:36 < kartikdutt18Gitt> Also Sreenik, I think one of the issue regarding convolution layer not having groups will be solved. I currently have a PR for Depthwise convolution, I can make changes so that convolution accepts groups as a parameter rather than having a different layer.

16:37 < sreenik[m]> kartikdutt18[m]: That would be great

16:56 < PrinceGuptaGitte> Thanks sreenik @kartikdutt18

17:45 < rcurtin> well, my meeting got postponed, so I actually will be able to make the whole video meetup

17:50 < PrinceGuptaGitte> Great!

18:40 < rcurtin> to run an individual test suite:

18:40 < rcurtin> bin/mlpack_test -t NameOfTestSuite

18:40 < rcurtin> and to run an individual test case:

18:40 < rcurtin> bin/mlpack_test -t NameOfTestSuite/NameOfTestCase

18:47 < rcurtin> valgrind+gdb: https://tromey.com/blog/?p=731

19:25 < kartikdutt18Gitt> Thanks @rcurtin, I'll try it out.

21:11 < Manav-KumarGitte> Hello Everyone, I am Manav Kumar 3rd year computer science student. I have beginner level experience in ML and Deep Learning and want to participate in gsoc by working on this organizations one of the ideas ' Improvisation and Implementation of ANN Modules'. Can somebody guide me with it.

21:22 < zoq> Manav-KumarGitte: Hello, https://www.mlpack.org/community.html and https://www.mlpack.org/gsoc.html should help you get started.

21:38 < zoq> saksham189Gitter: About adaptive pooling, sounds like a good idea to me, ideally we can avoid code-duplication, because each line has to be maintained.

23:10 < shrit[m]> Anyone knows if armadillo iterators has value_type traits?? I was not able to find this traits for the iterators