00:53 < sumedhghaisas__> zoq: Hey Marcus... So I tried different frameworks

00:54 < sumedhghaisas__> but now I know why he picked this one... so its not a classical classification task...

00:54 < sumedhghaisas__> cause the next symbol prediction probabilistic...

00:55 < sumedhghaisas__> thus... 7 separate classical classification task done in 1 network

00:55 < sumedhghaisas__> rcurtin: Hey Ryan...

00:55 < sumedhghaisas__> have some time?

00:56 < sumedhghaisas__> anyways... the test is passing now... but I did not change a lot. Don't know why it is passing now.

00:56 < sumedhghaisas__> ahh shit... I realized now.. the test was passing all along...

00:57 < sumedhghaisas__> I forgot to set the Rho for test cases

00:57 < sumedhghaisas__> I did that this time

00:57 < sumedhghaisas__> so even for smaller sequences it was detecting for unrelated random extra characters

01:03 < rcurtin> sumedhghaisas__: sorry, I am about to go to bed really early... I am traveling to Germany tomorrow for a conference so I am trying to get a head start with the jet lag :)

01:03 < rcurtin> I'll be back around 10am UTC

01:03 < rcurtin> ...if I can get up to the alarm :)

01:03 < sumedhghaisas__> haha... no problem. Will talk to you tomorrow...

01:04 < sumedhghaisas__> problem is solved ... wanted to talk know your opinion on a different architecture. Can wait...

02:28 < sumedhghaisas__> zoq, rcurtin: Shifted LSTM similarly to support variable length. And it worked in the first try... cannot believe it.

10:17 < sgupta> rcurtin: I'm getting permission denied error while creating directories in /var/www/html. I want to put different tar balls of all Libraries and serve those files from there.

10:17 < sgupta> rcurtin: I meant from the Apache server, I want to put files and then download them inside my containers

10:19 < sgupta> rcurtin: what should I do?

10:26 < rcurtin> sgupta: have you taken a look at the permissions on the directory?

10:28 < sgupta> rcurtin: it's root

10:29 < rcurtin> root is the owner of the directory, I mean, what are the permissions on the directory? set them so that the user running the apache server (in this case www-data) can read the directory and files

10:36 < rcurtin> probably it would be better to mark /var/www/html/ as writable for user sgupta (I can add you to the www-data group so that that is possible)

10:36 < rcurtin> then avoid copying the files to /var/www/html/ as root

10:37 < rcurtin> but instead copy as sgupta, then make sure that the files are at least group-readable afterwards

10:38 < rcurtin> let me know if I can clarify anything, I can help out if you like

10:38 < sgupta> rcurtin: okay, I think I have to do a chmod -R g+r /var/www/html. Is it?

10:39 < rcurtin> well, it's a little more tricky than that for a couple reasons---

10:40 < rcurtin> for a directory to be readable, it has to have both r and x permissions

10:40 < rcurtin> and the other thing is that in this case, all of the files are owned by user root and group root

10:40 < rcurtin> so even if you give group read permissions, the user www-data (which is what apache runs as) won't be able to read the files, since it isn't a part of group 'root'

11:01 < zoq> rcurtin: When is your flight?

11:02 < rcurtin> leaves 5:30pm EDT, arrives ~8am CEST

11:03 < rcurtin> I see switzerland has a different power plug than the rest of europe, I guess I will need to buy another adapter

11:03 < rcurtin> since I don't think the adapter I have will fit in a grounded plug

11:05 < rcurtin> but tomorrow I am in frankfurt all day before I take the train south on Sunday, so that should give sufficient time to explore and find an adapter :)

11:06 < zoq> ah, right, I guess you just need an adapter for Type C, who needs the grounding pin anyway

11:07 < rcurtin> I have a type C adapter, but I am not sure it will fit into a type J socket

11:07 < rcurtin> the pinout is the same, I am just not sure if the geometry will work out based on pictures I've seen

11:07 < zoq> aren't you an electrical engineer? So all you need screwdriver right?

11:07 < rcurtin> :)

11:08 < rcurtin> if it got desperate I could probably make it work like that :)

13:42 < zoq> sumedhghais: Great, I'll take a look at the updates later today. Just glanced over it, can you remove the code lines you commented?

14:01 < kris1> Yes so we have a problem that out error rate is going down as mentioned in a example we our following but the classification accuray is very low compared what is given in the example

14:25 < zoq> kris1: on what dataset?

14:26 < kris1> The same dataset used by the example

14:26 < kris1> it’s sklearn example i sent earlier

14:27 < kris1> The full example is in the latest commit i sent in the rbm PR

14:57 < zoq> okay, where did you get the dataset? Because in the example they don't use the mnist dataset.

15:04 < zoq> kris: where did you get the dataset? Because in the example they don't use the mnist dataset.

16:42 < kris1> Well i just copied from the site_packages sklearn dataset

16:43 < kris1> and then i saved the train and test split in csv format if you want i can share it with you

16:43 < zoq> kris1: yeah that would be helpful

16:51 < kris1> zoq: i am uploading the dataset now might take around 15 min of time

19:44 < rcurtin> sgupta: did you solve your permissions issue? let me know if you need help

21:41 < kris1> This might be a stupid question but does anybody know if can sync the random number generator of numpy and std::random using some seed values both using merssene twister for generation of the random number is it possible at all

22:05 < zoq> kris1: In theory, it should work.

22:10 < kris1> hmmm but i was not able to find any sources on google.

22:10 < zoq> kris1: I guess, math::RandomSeed(23); and random.seed(23)

22:10 < kris1> Ohkk i will have look at that

22:11 < kris1> Were you able to look at the sklearn example

22:11 < kris1> I told you about

22:14 < zoq> not sure numpy uses mt19937 as well

22:14 < zoq> Not completely yet, but I think Mikhail is right: https://github.com/mlpack/mlpack/pull/1027#pullrequestreview-46019659