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/
< hacsheng_>
I have a problem with building the mlpack Python bindings. I can use mlpack on CLI. When I open python and import mlpack, it shows no mlpack module.
< hacsheng_>
Does anyone encounter this problem before?
< rcurtin>
are the Python bindings available on your PYTHONPATH?
< rcurtin>
I would imagine they installed into /usr/local/lib/python.../ (or somewhere like this)
djhoulihan has quit [Quit: A deep and dreamless slumber.]
< hacsheng_>
I thought if I set "BUILD_PYTHON_BINDINGS=ON" before I make, Python bindings will be available...
< rcurtin>
yes
< rcurtin>
and if you build and install the Python bindings
< rcurtin>
you need to ensure that your PYTHONPATH includes the directory they were installed to
< rcurtin>
the first check would be to find the directory that the Python bindings installed to; you could search for mlpack/__init__.py somewhere on your system
djhoulihan has joined #mlpack
govg has quit [Ping timeout: 255 seconds]
govg has joined #mlpack
kaushik_ has quit [Quit: Connection closed for inactivity]
kaushik_ has joined #mlpack
< daivik>
rcurtin: Upgrading to boost 1.62 solved the problem. You were right, it was a version issue.
< rcurtin>
daivik: ok, that is good to know
< daivik>
also with v1.58, I can see 3 tests failing when i run mlpack_test -- and all are related to serialization
< rcurtin>
hmm, ok
< rcurtin>
so, if you are feeling up to it, you can test if this patch helps on boost 1.58:
< rcurtin>
unfortunately I can't dig in too deep, because the system with which I'd test boost 1.58 is about to go down for the weekend...
< rcurtin>
but anyway, the serialization issues I mentioned yesterday might be affecting some of those tests... but I am not sure. I'm kind of doubtful it will help your exact case
< rcurtin>
in any case, if you want to keep digging feel free; if not, definitely we should open a github issue at this point so that we can continue to track down the issue
< daivik>
Sure, I've got two docker containers set up now -- one with each boost version. I'll give it a go.
< rcurtin>
thanks, I'll be interested to see what you find
< rcurtin>
boost serialization is really nice but when something goes wrong it's a disaster ...