[mlpack-svn] [MLPACK] #300: allknn fails for mnist8m dataset
MLPACK Trac
trac at coffeetalk-1.cc.gatech.edu
Wed Nov 5 11:12:15 EST 2014
#300: allknn fails for mnist8m dataset
----------------------+-----------------------------------------------------
Reporter: rozyang | Owner: rcurtin
Type: defect | Status: accepted
Priority: major | Milestone:
Component: mlpack | Resolution:
Keywords: | Blocking:
Blocked By: 302 |
----------------------+-----------------------------------------------------
Comment (by rcurtin):
Wow -- 1TB... I don't have access to a machine with that much RAM. :(
Since I won't be able to reproduce the results, then, could I have you get
a little debugging output with gdb? If you can recompile mlpack with
debugging symbols (i.e. call CMake with the -DDEBUG=ON flag when you
configure mlpack), then run your program in gdb:
{{{
$ gdb allknn
...
(gdb) run -r mnist8m.csv -n mnist8m_3nn.csv -d distances_out.csv -k 3 -v
}}}
When it crashes, if you can get a backtrace (bt), that would probably be
helpful enough to figure out what is going on.
Also, do you know if ARMA_64BIT_WORD is enabled in Armadillo? You can do
that by modifying include/armadillo_bits/config.hpp, or adding `#define
ARMA_64BIT_WORD` at the top of your program (before `#include
<mlpack/core.hpp>`). If it isn't enabled, then Armadillo will refuse to
create very large matrices because they cannot be indexed uniquely with a
32-bit type.
--
Ticket URL: <http://trac.research.cc.gatech.edu/fastlab/ticket/300#comment:7>
MLPACK <www.fast-lab.org>
MLPACK is an intuitive, fast, and scalable C++ machine learning library developed at Georgia Tech.
More information about the mlpack-svn
mailing list