04:37 < AbishaiEbenezerG> hi mlpack! i do not know how to run the rl implementations and test cases with gym. could someone help me out?

04:37 < AbishaiEbenezerG> i went through https://github.com/zoq/gym_tcp_api#getting-started but i'm not sure if this is what i'm looking for

04:38 < AbishaiEbenezerG> i remember a conversation earlier here that said that we don't have to use this api anymore since mlpack already has the envs...something like that

04:39 < AbishaiEbenezerG> and especially since there isn't any documentation on this....

04:42 < AbishaiEbenezerG> even #1533 is kinda unclear to me...

05:07 < AbishaiEbenezerG> @sriramsk1999 need your help...

05:11 < SriramSKGitter[m> Hi @abinezer what exactly are you having troubles with?

05:12 < AbishaiEbenezerG> hey! so, i'm just starting out with trying out the rl algos with gym

05:13 < AbishaiEbenezerG> i'm a beginner here, so i might ask a few dumb stuff

05:14 < SriramSKGitter[m> To learn the ropes, I think it would be better to experiment with mlpack's inbuilt environments before trying out openai gym

05:14 < AbishaiEbenezerG> so lets say for example if i wanna try out the mountain car env with mlpacks q_learning

05:14 < AbishaiEbenezerG> how do i do that?

05:15 < SriramSKGitter[m> #2295 is a work in progress, but I think it should be of some help to you :)

05:16 < AbishaiEbenezerG> oh cool thanks ! i'll check it out rn and get back back if i have any more probs...

05:16 < SriramSKGitter[m> For more details, look at some of the tests like `q_learning_test.cpp`, `rl_components_test.cpp` and `async_learning_test.cpp`

05:17 < SriramSKGitter[m> No problem :)

05:17 < AbishaiEbenezerG> sure

07:14 < jenkins-mlpack2> Project docker mlpack nightly build build #640: STILL FAILING in 3 hr 0 min: http://ci.mlpack.org/job/docker%20mlpack%20nightly%20build/640/

07:18 < JoelJosephGitter> @abinezer https://codeshare.io/2BLeVb - this is some code i used to do mountaincar task, inspired from the "q_learning_test.cpp".... hope it helps :)

07:26 < JoelJosephGitter> Does the ffn.hpp have a backward function which can accept a custom loss function?

07:27 < JoelJosephGitter> backward function = for doing backpropagation

07:28 < JoelJosephGitter> i can see that pytorch has that feature, if im right

07:46 < pickle-rick[m]> How might I go about debugging a network that's not converging? In my case, I'm testing a vgg16 implementation(it compiles but trains slow and doesn't converge) , and I'm not sure what's going wrong. You could checkout #62 in mlpack/models repo for the code. Thanks in advance.

08:12 < kartikdutt18Gitt> Hey @adithya-tp, I have left some comments on your PR. Hope they help. Thanks.

08:48 < AbishaiEbenezerG> sure @joeljosephjin , i'll have a look at it....

09:14 < AbishaiEbenezerG> @sriramsk1999 where do i find this? \ref mlpack::rl

09:14 < AbishaiEbenezerG> "complete API documentation".

09:14 < AbishaiEbenezerG> For further documentation on the rl classes, consult the \ref mlpack::rl

09:14 < AbishaiEbenezerG> "complete API documentation".

09:23 < SriramSKGitter[m> @abinezer : https://www.mlpack.org/doc/mlpack-3.2.2/doxygen/namespacemlpack_1_1rl.html

09:52 < himanshu_pathak[> Zoq: I have completed the implementation RBFN locally by the I will push all changes with one test but there are many things left such as implementing multiple activation functions can I include this work in my proposal or not and yes implementation is quite slow because I am using for loop for some operation I will try to use arma functions for operations and improve it lot of work is left I thought it will be quite

09:52 < himanshu_pathak[> easy for me to implement I will complete it in a week or so but no it's taking much time. Sorry for delay I will update my pr by this evening

09:59 < pickle-rick[m]> > Hey @adithya-tp, I have left some comments on your PR. Hope they help. Thanks.

09:59 < pickle-rick[m]> Thanks!

10:00 < jeffin143[m]> @himanshu_pathak:matrix.org: it's totally ok , it's not everytime we can follow up with the expected plan , so being liltle late is ok , You shouldn't be sorry about it :)

10:31 < himanshu_pathak[> Yes jeffin143 Sometimes things does not happen as planned next time I will try to stick to the timeline

11:07 < AbishaiEbenezerG> @joeljosephjin

11:07 < AbishaiEbenezerG> while compiling i am getting this error

11:09 < AbishaiEbenezerG> i didn't understand the first error - the parameters passed looked fine

11:09 < AbishaiEbenezerG> did this same code work for you?

11:10 < AbishaiEbenezerG> error: ‘cout’ was not declared in this scope

11:10 < AbishaiEbenezerG> 63 | cout << "Average return: " << averageReturn.mean()

11:11 < AbishaiEbenezerG> the above 2 lines is part of the error as shown above. i don't know whats wrong here. i even added using namespace std. not sure why these basic errors are showing up, but i'll continue looking at this

14:00 < jeffin143[m]> zoq , rcurtin what happened to azure builds , can't see them in checks

14:33 < jeffin143[m]> rcurtin : you there ?

14:52 < nishantkr18[m]> <rcurtin "okay, strange, in this case I am"> rcurtin: I finally installed and ran with -lmlpack, so everything works now.. but still if i try to make it run using -L instead of -lmlpack, its giving `undefined referrence` error.

15:17 < nishantkr18[m]> <nishantkr18[m] "rcurtin: I finally installed and"> Sorry I couldn't get u before, as u had mentioned, I just found out that -lmlpack is also required to search for the library mlpack, along with -L, finally works without installing, sorry for the confusion.. feeling so stupid 😶

16:54 < zoq> jeffin143[m]: Not sure, will take a look into the issue later.

17:00 < zoq> jeffin143[m]: https://status.dev.azure.com/

17:00 < zoq> jeffin143[m]: https://status.dev.azure.com/_event/179641421

17:01 < zoq> I guess, we just have to wait.

17:02 < jeffin143[m]> Ok

17:02 < jeffin143[m]> That was quick

17:02 < jeffin143[m]> They must be really tensed , and working over time to solve this issue 😂

18:39 < favre49> A lot of olleges are closing because of COVID-19. I hope everyone stays safe and healthy :)

18:39 < favre49> On the plus side, I guess I have a lot of time to work on mlpack

18:47 < jeffin143[m]> So true , work from home in my office

18:47 < jeffin143[m]> Have spare time for Mlpack :)

18:48 < jeffin143[m]> And also we will get more PR , now since everyone will be working 😉

19:04 < GauravSinghGitte> Hi, @zoq shouldn't there be a fatal error if `MaxIterations()` parameter of an optimizer is defined as negative?

19:05 < himanshu_pathak[> I don't think Gaurav Singh (Gitter): anyone is going to do that.

19:30 < GauravSinghGitte> Hi @himanshupathak21061998 I also think that but when I did that just to see what happens with the `progress_bar`callback I got this result

19:50 < PrinceGuptaGitte> Hi @jeffin143, can you have a look at PR #2148 and #2192. I think they are ready from my side but any further suggestion to make improvements are appreciated.

19:50 < PrinceGuptaGitte> Thanks.

19:50 < PrinceGuptaGitte> (edited) Thanks. => Hi @jeffin143, can you have a look at PR #2184 and #2192. I think they are ready from my side but any further suggestion to make improvements are appreciated.

19:51 < PrinceGuptaGitte> They've been ready for some time now. I've also done the refactoring to avoid r value references.

19:52 < PrinceGuptaGitte> (edited) They've been ready for some time now. I've also done the refactoring to avoid r value references. => Hi @jeffin143, can you have a look at PR #2184 and #2192 if you get time. I think they are ready from my side but any further suggestion to make improvements are appreciated.

20:05 < favre49> It seems that the function definition is not showing at https://www.mlpack.org/doc/mlpack-3.2.2/doxygen/classmlpack_1_1ann_1_1TanhFunction.html#details

20:08 < favre49> Seems to be an issue with HardSigmoid as well

20:11 < favre49> The GELU and LiSHT functions are missing entirely from the docs

20:13 < favre49> doxygen function definitions are broken across the board, from what I can tell. I don't know enough to debug this right now (also need to go to sleep), so just letting everyone know

20:14 < favre49> Ah wait actually it makes sense classes are missing, since this is documenting a release. Seems like I really need some sleep :)

20:15 < favre49> Good night, everyone

21:08 < rcurtin> favre49: :), they should be showing for the mlpack-git documentation

21:58 < togo> what is the support level for Nvidia Tesla cards with this machine learning library?

22:17 < himanshu_pathak[> zoq (@freenode_zoq:matrix.org): Hey zoq when you get time please look at this pr https://github.com/mlpack/mlpack/pull/2261 right now the thyroid test is working and giving 0.07 classificationerror I will try to make mnist test also work right now it is not if you have any suggestions please provide.