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/
< kris1>
lozhnikov: i made the changes and i compared with both the scikit learn and deep learning frame work. I even added support for batch but still the training is not going well could have a look at it
< lozhnikov>
kris1: ok, I'll take a look
< kris1>
Thanks, the unit test can only be written once the rbm is trained. So i think i will write them once we figure out how the training part
< sgupta>
rcurtin: I'm trying to build boost from source. So, when run bootstrap.sh and do all the steps, being the root user, I'm able to build mlpack. But if I change my user to non root, I get a cmake error (unable to find requested boost libraries) while building mlpack.
< sgupta>
rcurtin: boost is installed to /usr. Any clues?
vivekp has joined #mlpack
vivekp has quit [Ping timeout: 255 seconds]
< rcurtin>
sgupta: can you show me both Dockerfiles and the error output?
< sgupta>
rcurtin: I'm in a cab right now. Will mail you those.
< rcurtin>
sure, sounds good
vivekp has joined #mlpack
vivekp has quit [Ping timeout: 246 seconds]
vivekp has joined #mlpack
sgupta has quit [Ping timeout: 260 seconds]
sumedhghaisas has joined #mlpack
sgupta has joined #mlpack
shikhar has joined #mlpack
kris1 has joined #mlpack
shikhar has quit [Quit: WeeChat 1.7]
kris1 has quit [Quit: kris1]
kris1 has joined #mlpack
mentekid has joined #mlpack
vivekp has quit [Ping timeout: 268 seconds]
vivekp has joined #mlpack
< ironstark>
virtual memory exhausting while building mlpack
< ironstark>
i am using the install_all.sh script provided in the benchmarks folder
< ironstark>
*libraries folder
< rcurtin>
how much RAM do you have?
< zoq>
and did you configure swap?
< ironstark>
I have 4GB RAM
< ironstark>
on my system
< ironstark>
and 8 GB swap space
< rcurtin>
I don't think mlpack will exhaust 4GB RAM during compilation, are you sure you are not running too many other processes?
< rcurtin>
also you can reduce -j4 to lower
< rcurtin>
so run install_all.sh 2 instead of install_all.sh 4
< ironstark>
okay
< ironstark>
thanks :)
< rcurtin>
yeah, that should fix the problem... if not, let us know and we can try more things
kartik has joined #mlpack
< ironstark>
sure
< kartik>
hi <zoq>
< kartik>
i took a day off ..
< kartik>
if things were working fine im planning to again make the changes from the same place when it was working ..
< kartik>
would take me a day but i think i can fix this
< kartik>
what do u say?
< zoq>
kartik: Hello, okay, it would be helpful if you could write the test I talked about on the mail and on the PR.
< rcurtin>
ironstark: right, I saw those, I am curious to check that the final centroids in milk and number of iterations are the same as mlpack for the given initial centroids
< kartik>
oh that one .. ohkae ..
< kartik>
ill give it now
< rcurtin>
it looks about right based on the benchmarking numbers you posted, but we can only know if it's exactly the same either by manual inspection or looking at the metrics output (assuming the metrics are outputting the number of iterations)
< kartik>
thanks zoq
vivekp has quit [Ping timeout: 260 seconds]
< rcurtin>
ironstark: I am looking forward to Shikhar's k-means changes being merged, I think that will improve the mlpack numbers :)
< ironstark>
rcurtin: How can we take a look into that?
< zoq>
kartik: Sure, once we have the test, I'll go and make some hopefully useful comments :)
< rcurtin>
ironstark: you could run the Python commands by hand for mlpack and for milk, and just compare the numpy centroids arrays you get back
< ironstark>
okay
< rcurtin>
or if RunMetrics() is collecting the number of iterations, then you can just run for a few datasets and make sure they are the same for both libraries
< rcurtin>
they'll either be the same or off by one (depending on how each library counts)
< ironstark>
okay I'll check and post the numbets
< ironstark>
*numbers
< rcurtin>
sounds good, thanks
vivekp has joined #mlpack
mikeling is now known as mikeling|afk
aashay has quit [Quit: Connection closed for inactivity]
aashay has joined #mlpack
aashay has quit [Changing host]
aashay has joined #mlpack
< rcurtin>
sgupta: I realized that jenkins@dealgood isn't a part of the docker group, hence the docker registry test #10 build failure
< rcurtin>
I fixed that and restarted docker on that system
mikeling|afk has quit [Quit: Connection closed for inactivity]