verne.freenode.net changed the topic of #mlpack to: http://www.mlpack.org/ -- We don't respond instantly... but we will respond. Give it a few minutes. Or hours. -- Channel logs: http://www.mlpack.org/irc/
killer_bee[m] has quit [Ping timeout: 252 seconds]
killer_bee[m] has joined #mlpack
killer_bee[m] has quit [Ping timeout: 256 seconds]
daivik has joined #mlpack
daivik has quit [Client Quit]
daivik has joined #mlpack
daivik has quit [Quit: http://www.kiwiirc.com/ - A hand crafted IRC client]
killer_bee[m] has joined #mlpack
daivik has joined #mlpack
daivik has quit [Quit: http://www.kiwiirc.com/ - A hand crafted IRC client]
vivekp has quit [Ping timeout: 240 seconds]
vivekp has joined #mlpack
daivik has joined #mlpack
alsc has joined #mlpack
daivik has quit [Quit: http://www.kiwiirc.com/ - A hand crafted IRC client]
vivekp has quit [Ping timeout: 240 seconds]
vivekp has joined #mlpack
govg has quit [Ping timeout: 260 seconds]
govg has joined #mlpack
djhoulihan has joined #mlpack
daivik has joined #mlpack
daivik has quit [Quit: http://www.kiwiirc.com/ - A hand crafted IRC client]
ImQ009 has joined #mlpack
daivik has joined #mlpack
travis-ci has joined #mlpack
< travis-ci> mlpack/mlpack#3877 (master - 8403b92 : Ryan Curtin): The build has errored.
travis-ci has left #mlpack []
< rcurtin> ok, I will try rebooting masterblaster now...
jenkins-mlpack has quit []
jenkins-mlpack has joined #mlpack
< zoq> looks like everything works
daivik has quit [Quit: http://www.kiwiirc.com/ - A hand crafted IRC client]
travis-ci has joined #mlpack
< travis-ci> ShikharJ/mlpack#60 (Test - dae4106 : Shikhar Jaiswal): The build has errored.
travis-ci has left #mlpack []
< rcurtin> yep, good to see :)
travis-ci has joined #mlpack
< travis-ci> ShikharJ/mlpack#61 (Test - a6e4443 : Shikhar Jaiswal): The build has errored.
travis-ci has left #mlpack []
alsc has quit [Quit: alsc]
daivik_ has joined #mlpack
ShikharJ has joined #mlpack
< ShikharJ> rcurtin: I am still getting the build errors, inspite of removing CmakeCache.txt file and also the build/ directory. I'm practically unable to test my code locally, and am pushing blindly to Travis. What else can be done?
< ShikharJ> The same log: Building CXX object src/mlpack/methods/sparse_coding/CMakeFiles/mlpack_sparse_coding.dir/sparse_coding_main.cpp.o [ 45%] Linking CXX executable ../../../../bin/mlpack_sparse_coding [ 45%] Built target mlpack_sparse_coding [ 45%] Generating CXX prefix source src/mlpack/tests/cotire/mlpack_test_CXX_prefix.cxx [ 45%] Generating CXX prefix header src/mlpack/tests/cotire/mlpack_test_CXX_prefix.hxx make[2]: *** No rule to ma
< rcurtin> the message was cut off, is it still about cli_deleter.hpp?
< ShikharJ> make[2]: *** No rule to make target 'src/mlpack/core/util/cli_deleter.hpp', needed by 'src/mlpack/tests/cotire/mlpack_test_CXX_prefix.hxx.gch'. Stop. CMakeFiles/Makefile2:4371: recipe for target 'src/mlpack/tests/CMakeFiles/mlpack_test.dir/all' failed make[1]: *** [src/mlpack/tests/CMakeFiles/mlpack_test.dir/all] Error 2 Makefile:140: recipe for target 'all' failed make: *** [all] Error 2
< ShikharJ> Yes
< rcurtin> so you have entirely new build directory and you are still getting this problem?
< ShikharJ> Yep. I tried first by just removing the CmakeCache.txt file, didn't work, then removed the build directory as well.
< rcurtin> are you working in a branch? is the branch synced to upstream master?
< rcurtin> and does the phrase 'cli_deleter' show up anywhere in the repository src/ directory?
< rcurtin> i.e. 'grep -r cli_deleter src/'
< ShikharJ> It is rebased to the latest master, yes.
< daivik_> I was facing the same problem as ShikharJ, but deleting and recreating the build directory, reconfiguring and running make solved the issue
< ShikharJ> It gives a vast log. Something like this: grep -r cli_deleter src/ src/mlpack/bindings/python/tests/CMakeFiles/generate_pyx_test_python_binding.dir/depend.make:src/mlpack/bindings/python/tests/CMakeFiles/generate_pyx_test_python_binding.dir/__/generate_pyx_test_python_binding.cpp.o: src/mlpack/core/util/cli_deleter.hpp src/mlpack/bindings/python/tests/CMakeFiles/generate_pyx_test_python_binding.dir/depend.make:src/mlpack/bindings/p
< rcurtin> no, run the grep in the source directory, not the build/src/ directory
< ShikharJ> I ran it in the main folder, but now it doesn't show anything.
< rcurtin> ok, are you sure that you deleted your build folder and reconfigured?
< rcurtin> because this looks to me like there are still old configuration artifacts in your build folder
< ShikharJ> I deleted the build folder for sure, I'll try again and update it here.
< rcurtin> ok, I think that should fix it if you remove the build folder entirely and start over
travis-ci has joined #mlpack
< travis-ci> mlpack/mlpack#3880 (master - 01eae23 : Ryan Curtin): The build has errored.
travis-ci has left #mlpack []
< daivik_> rcurtin: I am trying some of the HMM command line tools. I was able to train an HMM (using hmm_train) and store the model in an output model file (xml). However, when I input the same model file to hmm_loglik, it gives the following error "DiscreteDistribution::Probability(): received observation 0; observation must be in [0, 0] for this distributi
< daivik_> on." I have double checked that the sequence I'm trying to compute the loglikelihood for has only those labels that are present in the training file. Any clue on what I'm doing wrong?
< rcurtin> hmm, maybe you have found a bug, but before we are sure, can you tell me what the sequence you are passing to hmm_loglik is?
alsc has joined #mlpack
< rcurtin> and the data you are passing to hmm_train also?
< daivik_> https://thepasteb.in/p/lOhOKAvXW66FB --> training data to hmm_train
< daivik_> 0
< daivik_> 2
< daivik_> 1
< daivik_> to hmm_loglik
travis-ci has joined #mlpack
< travis-ci> ShikharJ/mlpack#61 (Test - a6e4443 : Shikhar Jaiswal): The build has errored.
travis-ci has left #mlpack []
< rcurtin> if you saved the HMM as an XML, can you paste that too?
< rcurtin> the data is looking correct to me
< daivik_> https://thepasteb.in/p/O7h5X0ywPPZiq --> model file in xml format
< rcurtin> perfect, thank you
< rcurtin> so what I am looking for here is the emission probabilities for each hidden state
< rcurtin> "<probabilities ...>"
< rcurtin> each of those should be a single-dimension distribution with 12 elements in them
< rcurtin> 12 or 13, not sure of the exact number, but anyway something like that
< daivik_> 12
< rcurtin> right
< daivik_> I can confirm that the model file is correct -- and is being written (by hmm_train) and read (by hmm_loglik) correctly
< rcurtin> yeah, it seems correct, so the question is why DiscreteDistribution::Probability() is having problems
< rcurtin> are you willing to pull out the debugger and trace what is going on? I can help point you in the right direction
< daivik_> yeah sure
< daivik_> you'll have to tell me the steps though (sorry for being a noob)
< rcurtin> it's no problem; honestly printf-style debugging can get us most of what we need here
< rcurtin> the file I'm interested in is discrete_distribution.cpp in src/mlpack/core/dists/
< rcurtin> our failure is on line 85
< rcurtin> I would say, inside of that if statement, add the following output:
< rcurtin> std::cout << "number of dimensions in DiscreteDistribution: " << probabilities.size();
< rcurtin> std::cout << "; input observation:\n" << observation;
< rcurtin> ack, that should be "<< observations;"
< rcurtin> std::cout << "probabilities for dimension " << i << ": " << probabilities[i].t();
< rcurtin> std::cout << "obs: " << obs << "\n";
< rcurtin> and I think that should give some interesting output that will help figure out what is going on
< daivik_> okay .. let me try that and I'll get back to you here in a few minutes
< rcurtin> sure, I hope there are no errors in what I wrote :)
< rcurtin> but ideally that should give us some output on what is actually going on inside the DiscreteDistribution
< daivik_> no output, a similar error is printed by line 144 -- should i add the cout statements there as well
< daivik_> I think the correct place to put the cout is in discrete_distribution.hpp line 146 .. rather than where we did
< daivik_> because that is the function mentioned in the error message
< rcurtin> ah yeah, sorry
< rcurtin> my mistake
< rcurtin> yeah, I was thinking line 144 but for some reason gave the wrong line number
travis-ci has joined #mlpack
< travis-ci> mlpack/mlpack#3882 (master - eae2453 : Ryan Curtin): The build has errored.
travis-ci has left #mlpack []
kaushik_ has joined #mlpack
alsc has left #mlpack []
ShikharJ has quit [Quit: Page closed]
< daivik_> number of dimensions in DiscreteDistribution: 1; input observation:
< daivik_> 0
< daivik_> probabilities for dimension 0: [matrix size: 1x0]
< daivik_> obs: 0
< rcurtin> well that is not expected
< rcurtin> let me see if I can reproduce this
< rcurtin> or would you like to continue debugging?
< daivik_> These are the lines I added to discrete_distribution.hpp (rather than discrete_distribution.cpp)
< daivik_> std::cout << "number of dimensions in DiscreteDistribution: " << probabilities.size();
< daivik_> std::cout << "; input observation:\n" << observation;
< daivik_> std::cout << "probabilities for dimension " << dimension << ": " << probabilities[dimension].t();
< daivik_> std::cout << "obs: " << obs << "\n";
< rcurtin> basically we now know that there is some DiscreteDistribution that has 0 dimensions
< daivik_> I think I'd like to work on this a bit more
< rcurtin> but now the next question is, why is there a DiscreteDistribution with 0 dimensions?
< rcurtin> sure, I can keep providing pointers if you like :)
< daivik_> yeah, i had sort of arrived at the same conclusion too
< daivik_> i'll ask here if I get stuck if thats okay?
< rcurtin> sure, of course
< daivik_> thanks a lot for your help
< rcurtin> I think maybe the next place to look would be to print the HMM emission probabilities in HMM::Serialize()
< rcurtin> after the HMM is loaded
< daivik_> okay, I'll have a look at that
< rcurtin> sorry, HMM::serialize(), hmm_impl.hpp:590
< rcurtin> after line 603, the emissions should be loaded, and you could print the emission distribution probabilities there
< daivik_> I was hoping to write tests for the hmm command line bindings -- but I think I'll try and solve this first :)
< rcurtin> yeah, I think for many binding tests, some bugs in the underlying code get discovered :)
daivik_ has quit [Quit: http://www.kiwiirc.com/ - A hand crafted IRC client]
daivik_ has joined #mlpack
< rcurtin> ok, I'll be shutting down all the mlpack build servers in Oregon at 1800 UTC on Friday (masterblaster, gekko, scrooloos, savannah, slake, skyfish, aunty, collector, ironbar) and they should be received in Tucson Arizona on Monday and re-racked shortly thereafter
< zoq> fingers crossed
ImQ009 has quit [Quit: Leaving]
daivik_ has quit [Quit: http://www.kiwiirc.com/ - A hand crafted IRC client]
travis-ci has joined #mlpack
< travis-ci> ShikharJ/mlpack#63 (RBM - f47f3e7 : Shikhar Jaiswal): The build has errored.
travis-ci has left #mlpack []
kaushik_ has quit [Quit: Connection closed for inactivity]
travis-ci has joined #mlpack
< travis-ci> ShikharJ/mlpack#62 (master - eae2453 : Ryan Curtin): The build has errored.
travis-ci has left #mlpack []
travis-ci has joined #mlpack
< travis-ci> ShikharJ/mlpack#64 (RBM - 92a29c3 : Shikhar Jaiswal): The build has errored.
travis-ci has left #mlpack []