04:04 < favre49> I've been out of the loop, there haven't been any IRC logs since 21st of September?

04:09 < favre49> Ah the funny thing is if i put off my laptop I'll never see the responses to this :)

04:11 < favre49> I'm not sure if you guys already know, but because of Hacktoberfest lots of repos have been inundated with spam. Doesn't seem like those people have found us yet, but wanted to give a heads up in case this happens here as well. I think github now has the option to limit interactions based on their role in the organization

04:37 < jeffin143[m]> favre49: oss community is on a rant against hacktoberfest

04:37 < jeffin143[m]> Just pure rant

04:38 < jeffin143[m]> And the main culprits are Indians

04:39 < jeffin143[m]> Lots of spammy pr from numerous indians , it hasn't been so bad till now , But this year it went out of hands

04:42 < favre49> Well, an Indian youtuber made a video about how to "get free t shirts easily". Poor decision on his part, he basically created a spam bot. This was inevitable with Hacktoberfest's way of working though.

04:46 < jeffin143[m]> Ahaaaa I see ? Was it some amazing project related ? Because it was 131290 , when I search amazing project keyword

04:46 < jeffin143[m]> In 7 hrs

04:46 < jeffin143[m]> That was huge

04:47 < jeffin143[m]> Every one was adding amazing project word to readme , and what not

05:36 < AakashkaushikGi4> I read on the link that say4n sent and the screenshots there were almost all indian people, and i was reading the discussion here and searched YouTube randomly with the tag hacktoberfest and there are so many videos to complete hacktoberfest in 5-10 min, I don't really understand why people even try to spam out such things when there are easy issues in a lot of repositories they can help with.

07:14 < say4n> Here's more on the situation: https://joel.net/how-one-guy-ruined-hacktoberfest2020-drama :(

08:16 < RishabhGarg108Gi> Hello Everyone! I was trying to build the mlpack from its source code. In the documentation, it is mentioned to download mlpack-3.0.4. So, my doubt is shouldn't we clone the actual repository from github to work on it ? And is that specified version to download in the docs is the lastest one ?

09:03 < AakashkaushikGi4> @say4n yeah it's just a lot of pain for the maintainers.

09:04 < AakashkaushikGi4> @say4n it's just a lot of pain for the maintainers.

09:07 < AakashkaushikGi4> Hey @RishabhGarg108 can you send a link to the documentation to which you are referring to and the latest stable version of mlpack is 3.4.1

09:32 < RishabhGarg108Gi> Hey @Aakash-kaushik, https://www.mlpack.org/doc/mlpack-3.1.0/doxygen/build.html this is the link to the docs which i am referring.

09:35 < RishabhGarg108Gi> Could you also tell me which is the latest development version on which I could work. Thanks

10:10 < zoq> RishabhGarg108Gi: The latest dev version would be the git master branch - https://github.com/mlpack/mlpack

10:13 < zoq> RishabhGarg108Gi: Where does the link come from? Because the website should link to the latest stable version which is 3.4.1.

10:14 < RishabhGarg108Gi> I reached there from this https://www.mlpack.org/community.html

10:14 < RishabhGarg108Gi> It has a link called build mlpack from source

10:14 < zoq> RishabhGarg108Gi: I see thanks.

10:15 < zoq> RishabhGarg108Gi: Will update the link.

10:15 < zoq> RishabhGarg108Gi: The install steps are still the same, so no problem there.

10:16 < RishabhGarg108Gi> Okay. Thanks a lot @zoq

11:50 < yashwants19[m]> Hey [zoq](https://matrix.to/#/@freenode_zoq:matrix.org) we are still facing the same problem for auto-cancel.

11:50 < yashwants19[m]> But cache is working great.

11:52 < yashwants19[m]> Do this problem belong to some internal action settings?

11:52 < yashwants19[m]> And should we create a personal access token?

11:52 < zoq> yashwants19[m]: Just checked Actions permissions which is set to "Allow all actions".

11:53 < zoq> yashwants19[m]: Good idea.

11:53 < yashwants19[m]> Not sure why this is happening.

11:54 < zoq> yashwants19[m]: Might be a good idea to test this on a fresh repo?

11:54 < zoq> yashwants19[m]: Or does ths work on your fork?

11:55 < yashwants19[m]> Yeah this sounds good

11:55 < yashwants19[m]> <zoq "yashwants19: Or does ths work on"> Yup

11:55 < yashwants19[m]> It was working fine on my repo.

11:55 < yashwants19[m]> I mean fork.

11:56 < zoq> Okay, let me open a PR and see if it works for me.

11:56 < yashwants19[m]> Awesome.

11:58 < yashwants19[m]> It is always trying to cancel the builds, but faces that 'resource not accessible by integration' issue.

12:00 < zoq> yashwants19[m]: https://github.com/mlpack/mlpack/pull/2649

12:00 < zoq> yashwants19[m]: Same on my side.

12:01 < yashwants19[m]> Yes

12:01 < yashwants19[m]> May be you can give a try to personal access token

12:07 < zoq> yashwants19[m]: Do you know who enabled GitHub Actions?

12:07 < zoq> yashwants19[m]: https://docs.github.com/en/free-pro-team@latest/actions/reference/authentication-in-a-workflow mentions a token is automatically created, but I don't see the token.

12:08 < yashwants19[m]> I guess when [rcurtin](https://matrix.to/#/@freenode_rcurtin:matrix.org) merge the R-bindings, github actions enable after that merge.

12:10 < RishabhGarg108Gi> Hey, I ran the make command and it froze my computer at the following point and it did not proceed after that. Could someone help why it is freezing everything ?

12:12 < yashwants19[m]> cancel job is showing that token is provided https://github.com/mlpack/mlpack/runs/1198600442?check_suite_focus=true#step:3:16

12:12 < yashwants19[m]> https://github.com/mlpack/mlpack/runs/1198600442?check_suite_focus=true#step:3:4

12:13 < zoq> yashwants19[m]: Do you see a token in your repo?

12:13 < zoq> RishabhGarg108Gi: Do you build with make -j4 ?

12:14 < zoq> yashwants19[m]: Will take a closer look into the setup process later.

12:14 < RishabhGarg108Gi> yes @zoq

12:14 < zoq> RishabhGarg108Gi: Okay, try with make only without -j4

12:15 < yashwants19[m]> [zoq](https://matrix.to/#/@freenode_zoq:matrix.org) sure.

12:15 < zoq> RishabhGarg108Gi: -j4 will tel cmake to use multiple cores to build, which is fine as long as you have enough cores and memory.

12:15 < RishabhGarg108Gi> I tried that too. but that also froze at the same place.

12:15 < RishabhGarg108Gi> I have 8 codes and 8 Gb RAM

12:16 < zoq> RishabhGarg108Gi: It can take some time to build.

12:16 < RishabhGarg108Gi> Also I am not running any other application. Just a browser and terminal is running

12:16 < zoq> Or is the CPU usage also down?

12:18 < RishabhGarg108Gi> Actually everything runs fine until it reaches that 50% mark and after that it just freezes everything. I tried restarting my system 3 times and every time it just stucks there.

12:22 < zoq> RishabhGarg108Gi: Can you check in the build process it dosn't eat up all memory?

12:22 < zoq> yashwants19[m]: Okay, looks like - https://dev.to/dtinth/authenticating-as-a-github-app-in-a-github-actions-workflow-27co we have to create a new personal token.

12:23 < RishabhGarg108Gi> how can I do that ? I am running a linux machine.

12:24 < zoq> e.g. by keping an ey on htop

12:24 < zoq> eye

12:25 < AakashkaushikGi4> Hey @RishabhGarg108 i have a similar config and i generally build mlpack with make -j1

12:26 < AakashkaushikGi4> That should work fine for you too, it may take a bit long but it will build.

12:26 < AakashkaushikGi4> In terminal you can write `free -h` to check for the memory usage.

12:28 < AakashkaushikGi4> And if you are using something like ubuntu it has a system monitor, you can check that too.

12:29 < yashwants19[m]> [zoq](https://matrix.to/#/@freenode_zoq:matrix.org) I think so this might work.

12:29 < yashwants19[m]> https://github.com/styfle/cancel-workflow-action/issues/7

12:36 < zoq> yashwants19[m]: Wait we use 'styfle/cancel-workflow-action' ?

12:37 < yashwants19[m]> Yup

12:37 < zoq> yashwants19[m]: Okay, in this case we have to install the app :)

12:38 < yashwants19[m]> Github doesn't provide this.

12:38 < zoq> yashwants19[m]: Have to finish some other thing first, will install the app afterwards.

12:38 < yashwants19[m]> Awesome

12:39 < yashwants19[m]> Sounds good

12:39 < yashwants19[m]> Thank you for the help

12:43 < RishabhGarg108Gi> I monitored the resources using htop. And indeed it is consuming 100% of all the 8 cores and full 7.6 Gb RAM.

12:43 < himanshu_pathak[> Hey zoq: rcurtin I tried sklearn svm with rbf on mnist dataset and got accuracy of 93% which is quite near to my implementation 92.7 something but sklearn was too fast to train I need to do some changes in my Gaussian Kernel svm pr. Do you have any suggestion regarding this?

13:20 < rcurtin> himanshu_pathak[: I can try and take a look; the most important thing is to try and avoid temporary matrices in the computations

13:21 < rcurtin> one big place to start might be to try and profile to see where the slow part is

13:21 < rcurtin> you can use regular C++ benchmarking and profiling tools for this, and you could also use `Timer::Start()` and `Timer::Stop()` if you want to get a coarse-grained look (but know that there is some overhead when you call those functions)

13:21 < rcurtin> another thing to check is how many iterations your version takes in SMO vs. sklearn's---if they are converging way more quickly, then that will be a big runtime difference

13:30 < himanshu_pathak[> Ok rcurtin I will try these things and will share some results soon thanks for helping :)

13:36 < rcurtin> yeah, first I would check to see if you are converging in the same number of iterations, and if so, then we can start looking at why each individual iteration is slow

13:40 < AakashkaushikGi4> > I monitored the resources using htop. And indeed it is consuming 100% of all the 8 cores and full 7.6 Gb RAM.

13:40 < AakashkaushikGi4> were you finally able to build it with `make -j1` ?

15:31 < RishabhGarg108Gi> Hii @Aakash-kaushik , no it again ran froze everything. And above everything now my ubuntu has run into emergency mode and i am unable to get out of it.

16:20 < abernauer[m]> rcurtin: zoq Prioritizing those notebooks for the R bindings examples this weekend just a heads up.

16:50 < rcurtin> awesome, sounds good

17:29 < abernauer[m]> One quick dumb question. I need to update my versions of the R dependencies and my fork of mlpack.

17:30 < abernauer[m]> After doing that I should only need to run cmake with the R bindings argument and then run make to build bindings correct? My mlpack version was built from source.

17:30 < rcurtin> yeah, that should be correct; you'll then need to take the built R bindings and install them (I am less familiar on that part)

17:30 < rcurtin> you could look at the CI configuration for the R build, in .github/workflows/main.yml

17:52 < AakashkaushikGi4> @RishabhGarg108 I have never experienced such a issue myself of running into emergency on ubuntu, i agree it may hang or not respond but generally the issue of running into emergency (I don't exactly know what you mean)is not something that happens by building something, and the `j1` flag should have worked as it doesn't distributes the build on your cpu and so generally youe memory is not completely filled another

17:52 < AakashkaushikGi4> things you can try is increasing your swap space and then trying again with the same flag.

17:58 < RishabhGarg108Gi> Thanks @Aakash-kaushik , I will try to increase swap size. Actually I was looking on web of some way to limit the resource alllocated to a particular app/task. If it is possible then I could simply set it to some value at which my system does not freeze.

17:59 < RishabhGarg108Gi> Btw that emergency mode is now fixed :)

18:56 < AakashkaushikGi4> Hey, what is the use of the `delta` and the `outputParameter` parameters that are implemented in a lot of activation functions in ann/layer, what actually do these parameters store ?

19:52 < shrit[m]> rcurtin: I have tried to subscribe to mlpack-git from a week ago but I did not get any results.

20:23 < rcurtin> shrit[m]: yes, outlook.com is rejecting emails and I have no idea why; I've been trying to figure it out but the outlook team is not particularly helpful

20:23 < rcurtin> maybe subscribe with a different address? :(

20:24 < shrit[m]> Hmm, interesting I have subscribed with address of my domain name

20:24 < shrit[m]> shrit.me ??

20:25 < shrit[m]> If you can add me that will be perfect

20:25 < rcurtin> I can do that, but I can also see in the mailserver logs that emails are going to shrit.me and being successfully received by that mailserver

20:25 < rcurtin> maybe they are being marked as spam?

20:28 < rcurtin> anyway, I subscribed your shrit.me email, so hopefully it should start working

20:28 < rcurtin> the whole thing is kind of infuriating... I change IP, but SPF/DKIM/DMARC are all set up, yet still mailservers block my emails or send them to spam for no apparent reason, and nobody will tell me *why*

20:29 < rcurtin> outlook: "Our investigation has determined that the above IP(s) do not qualify for mitigation.

20:29 < rcurtin> "

20:29 < rcurtin> but... they don't say *why*!

20:30 < shrit[m]> They are not marked as spam, I am only receiving the github one.

20:30 < shrit[m]> Your configurations are fine, it is outlook that not implementing correclyl the RFC

20:36 < shrit[m]> rcurtin: I have just sent you an email with my discussion with their team for my domain name, if you can extract the email and try to contact them I think they will answer you eventually

20:42 < shrit[m]> I have just received the Welcome email thanks 👍️

20:43 < rcurtin> awesome, glad that worked at least

20:43 < rcurtin> that deliverability email you forwarded... Sai didn't say anything about why the block was in place

20:44 < rcurtin> but, in any case, if they do the same for me, great

20:44 < shrit[m]> Now, If I sent email to outlook it will go to junk, but at least it will not be filtered out by outlook

20:45 < shrit[m]> I know they did not specify, it seems as if they just blocked the entire IP in this world in order to force the people to use their email services

20:46 < rcurtin> I think you are right... :(

20:47 < shrit[m]> I am surprised that more and more institutions are using outlook exchange as their main email server.

20:48 < rcurtin> Georgia Tech did it so that they could reduce the cost of their paid IT staff

20:48 < rcurtin> apparently it is better to pay a contract to another company than it is to have people on the payroll

21:00 < shrit[m]> rcurtin: Just for info, since I am watching on Github, do I receive a double copy of emails now? or just one from the mlpack-git?

21:01 < rcurtin> yeah, you will receive a double-copy of *some* but not all emails

21:01 < rcurtin> personally I ignore the entire mlpack organization on github; I get everything through the list

21:02 < shrit[m]> Exactly.

21:03 < abernauer[m]> rcurtin: I think I just need to run R CMD INSTALL -l package from the location of the bindings. I will check the resource you mentioned earlier and I wouldn't be surprised if I have to clean up my R installs.