mlpack IRC logs, 2018-06-10

Logs for the day 2018-06-10 (starts at 0:00 UTC) are shown below.

>
June 2018
Sun
Mon
Tue
Wed
Thu
Fri
Sat
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
--- Log opened Sun Jun 10 00:00:57 2018
05:57 -!- ImQ009 [~ImQ009@unaffiliated/imq009] has joined #mlpack
06:43 -!- sds [9d26ce2c@gateway/web/freenode/ip.157.38.206.44] has joined #mlpack
06:43 -!- sds is now known as Guest65307
06:45 -!- Guest65307 [9d26ce2c@gateway/web/freenode/ip.157.38.206.44] has quit [Client Quit]
06:45 -!- vivekp [~vivek@unaffiliated/vivekp] has joined #mlpack
08:38 -!- wenhao [731bc65e@gateway/web/freenode/ip.115.27.198.94] has quit [Quit: Page closed]
08:40 < ShikharJ> zoq: I received some good results on the full dataset as well :) Posting in the PR now.
08:48 < ShikharJ> zoq: It took about 3 days for the GAN to converge, but the results we got were also better than what is posted as the output for the O'Reilly example here (https://github.com/mlpack/mlpack/pull/1066#issuecomment-322415665).
08:49 < ShikharJ> zoq: I think we can cut down a lot on the time elapsed, once the support for batches is implemented. We are all set to merge the GAN PR now!
08:50 < ShikharJ> zoq: Take a look at our outputs here (https://github.com/mlpack/mlpack/pull/1204#issuecomment-395187579).
09:11 -!- vivekp [~vivek@unaffiliated/vivekp] has quit [Ping timeout: 240 seconds]
09:14 -!- vivekp [~vivek@unaffiliated/vivekp] has joined #mlpack
09:37 < zoq> ShikharJ: The results are really promising, great.
09:37 -!- travis-ci [~travis-ci@ec2-54-205-136-216.compute-1.amazonaws.com] has joined #mlpack
09:37 < travis-ci> ShikharJ/mlpack#173 (GAN - 2635d6b : Shikhar Jaiswal): The build has errored.
09:37 < travis-ci> Change view : https://github.com/ShikharJ/mlpack/compare/9014d65def05...2635d6b5966f
09:37 < travis-ci> Build details : https://travis-ci.org/ShikharJ/mlpack/builds/390338785
09:37 -!- travis-ci [~travis-ci@ec2-54-205-136-216.compute-1.amazonaws.com] has left #mlpack []
09:49 < ShikharJ> zoq: Thanks for your help. I was able to debug our DCGAN MNIST test, I'll run the 10,000 image and 70,000 image tests on that as well.
10:18 < jenkins-mlpack> Project docker mlpack nightly build build #345: STILL UNSTABLE in 3 hr 4 min: http://masterblaster.mlpack.org/job/docker%20mlpack%20nightly%20build/345/
10:19 < ShikharJ> zoq: Do you think we can merge the GAN code now?
10:22 < zoq> ShikharJ: Yes let's merge the code, one last thing can you remove the output from the test (https://github.com/mlpack/mlpack/pull/1204/files#diff-7ad5b550c447d8de902a03acc4d31736R90)?
10:24 < ShikharJ> zoq: Do you mean the `std::cout << "Loading Parameters" << std::endl;` statement?
10:25 < zoq> yeah
10:25 < zoq> If you like keep the output, but we should use Log::Debug.
10:26 < zoq> That way a user can disable the output.
10:27 < ShikharJ> zoq: I'll make use of Log::Info?
10:27 < zoq> yeah, that's fine as well.
10:34 < ShikharJ> zoq: Done!
10:35 < zoq> ShikharJ: Okay, once the travis build finished, I'll hit the merge button :)
10:36 < ShikharJ> zoq: Great. COuld you also tell of a source where I can find the CelebA dataset in the csv format? The original author's link seems to have gone dead, and I'm only able to find the jpg or png format files.
10:43 < zoq> ShikharJ: Not sure the dataset exists as csv, but we could use the hdf5 format: https://github.com/tdeboissiere/DeepLearningImplementations/tree/master/GAN/src/data
10:43 < zoq> ShikharJ: Another idea is to create one as csv, I can do this if you like.
10:44 < ShikharJ> zoq: Does armadillo::load support hdf5 format?
10:45 < ShikharJ> zoq: Ah, I see now, it does. Thanks for that.
11:24 -!- travis-ci [~travis-ci@ec2-54-161-135-156.compute-1.amazonaws.com] has joined #mlpack
11:24 < travis-ci> ShikharJ/mlpack#174 (GAN - b294b31 : Shikhar Jaiswal): The build has errored.
11:24 < travis-ci> Change view : https://github.com/ShikharJ/mlpack/compare/2635d6b5966f...b294b31566cd
11:24 < travis-ci> Build details : https://travis-ci.org/ShikharJ/mlpack/builds/390357172
11:24 -!- travis-ci [~travis-ci@ec2-54-161-135-156.compute-1.amazonaws.com] has left #mlpack []
11:29 -!- vivekp [~vivek@unaffiliated/vivekp] has quit [Ping timeout: 248 seconds]
11:31 -!- vivekp [~vivek@unaffiliated/vivekp] has joined #mlpack
11:34 < ShikharJ> zoq: I've tmux'd the two DCGAN MNIST builds as well. I'll get to debugging the CelebA test as well.
12:01 -!- manish7294 [8ba7b34f@gateway/web/freenode/ip.139.167.179.79] has joined #mlpack
12:02 < zoq> ShikharJ: Good, I guess we could use a subset for the DCGAN as well to see some initial results.
12:03 -!- travis-ci [~travis-ci@ec2-54-161-135-156.compute-1.amazonaws.com] has joined #mlpack
12:03 < travis-ci> ShikharJ/mlpack#175 (DCGAN - 5a5451c : Shikhar Jaiswal): The build has errored.
12:03 < travis-ci> Change view : https://github.com/ShikharJ/mlpack/compare/9643d64a1881...5a5451c329c5
12:03 < travis-ci> Build details : https://travis-ci.org/ShikharJ/mlpack/builds/390358018
12:03 -!- travis-ci [~travis-ci@ec2-54-161-135-156.compute-1.amazonaws.com] has left #mlpack []
12:10 < manish7294> zoq: I was using BigBatchSGD with adaptive stepsize and it turns out that effectiveBatchSize is surpassing permissible value leading to errors, whereas backtracking line search works fine. Here is the backtrace https://pastebin.com/q0LWjM4V
12:13 < manish7294> initially I kept batch size of 50
12:58 -!- manish7294 [8ba7b34f@gateway/web/freenode/ip.139.167.179.79] has quit [Ping timeout: 260 seconds]
13:09 < Atharva> zoq: In ffn_impl.hpp line 188 : res += Evaluate(parameters, i, true); is their some reason to writing the third parameter as a boolean, because there is no definition of Evaluate with the third parameter as a boolean and I think it's just getting converted to size_t = 1? Am I missing something?
14:07 < ShikharJ> zoq: One build runs the 10,000 image subset, should be done in 8 hours from now.
14:44 < zoq> manish7294: Ahh, can I use the code from the PR to reproduce the issue?
14:44 < zoq> Atharva: There is an Evaluate function (https://github.com/mlpack/mlpack/blob/master/src/mlpack/methods/ann/ffn.hpp#L159) which takes a boolean as the last parameter; it's used to distinguish between training and testing.
14:45 < zoq> ShikharJ: Nice, let's see if we get some reasonable results.
14:45 < zoq> manish7294: Another optimizer you could test is SGRD with CyclicalDecay.
14:46 < Atharva> zoq: That's true but that function has four parameters, the one I pointed to has three.
14:46 < Atharva> The evaluate function with 3 parameters has size_t as the last parameter
14:52 < zoq> ohh, you are right, nice catch
14:52 < zoq> this should be Evaluate(parameters, i, 1, true);
14:53 < zoq> Do you like to open a PR or should I fix this?
14:53 < Atharva> It's just one line, if you can directly push it, it will be better.
14:54 < Atharva> Is it okay?
15:03 -!- sulan_ [~sulan_@563BE0E4.catv.pool.telekom.hu] has joined #mlpack
15:06 -!- manish7294 [8ba7b34f@gateway/web/freenode/ip.139.167.179.79] has joined #mlpack
15:09 < manish7294> zoq: I haven't pushed the change regarding optimizer on the PR yet because of the issue but you can reproduce it by just passing the BigBatchSGD through lmnn_main.cpp --- you will just have to make a very short change in the last part of lmnn_main.cpp
15:32 < zoq> Atharva: Yeah, no problem.
15:32 < manish7294> zoq: I tried SGDR. The results were similar to SGD and comes up with the same problem of coordinates matrix divergence we are facing with SGD. So, I guess BigBatchSGD and L-BFGS are the best for us.
15:32 < zoq> manish7294: Okay, I'll take a look into the issue.
15:33 < manish7294> zoq: great, Thanks for help.
16:13 -!- manish7294 [8ba7b34f@gateway/web/freenode/ip.139.167.179.79] has quit [Quit: Page closed]
16:41 < ShikharJ> zoq: You there/
16:41 < ShikharJ> ?
16:42 < zoq> ShikharJ: yes
16:42 < ShikharJ> I think we can close off the old PR and issues now. Like (https://github.com/mlpack/mlpack/pull/1066) and (https://github.com/mlpack/mlpack/issues/1206).
16:43 < zoq> agreed, let me close the issue/pr
16:46 < ShikharJ> zoq: For the next one week, I would like to focus on getting the support for batches and optimizer separation (and debug and test the DCGAN tests on the side). Would that be fine?
16:47 < zoq> ShikharJ: Absolutely, let us put some time into tuning the existing code.
17:06 -!- witness_ [uid10044@gateway/web/irccloud.com/x-xccspnlusqofaqkj] has joined #mlpack
17:14 -!- vivekp [~vivek@unaffiliated/vivekp] has quit [Ping timeout: 260 seconds]
20:36 -!- sulan_ [~sulan_@563BE0E4.catv.pool.telekom.hu] has quit [Quit: Leaving]
20:38 -!- ImQ009 [~ImQ009@unaffiliated/imq009] has quit [Quit: Leaving]
--- Log closed Mon Jun 11 00:00:58 2018