How can I solve Segmentation fault (core dumped) ?

[Jason Kim]

Hi,

I successfully installed gensim by 'pip install --upgrade gensim'.

But I've got segmentation fault when import module.

Some information is below.

root@digits-01:~# uname -v

#26~14.04.1-Ubuntu SMP Fri Jul 24 21:16:20 UTC 2015

root@digits-01:~# 

root@digits-01:~# python

Python 2.7.12 |Anaconda 4.1.1 (64-bit)| (default, Jul  2 2016, 17:42:40) 

[GCC 4.4.7 20120313 (Red Hat 4.4.7-1)] on linux2

Type "help", "copyright", "credits" or "license" for more information.

Anaconda is brought to you by Continuum Analytics.

Please check out: http://continuum.io/thanks and https://anaconda.org

>>> import numpy

>>> numpy.__version__

'1.11.1'

>>> import scipy

>>> scipy.__version__

'0.18.0'

>>> from gensim import utils

Segmentation fault (core dumped)

root@digits-01:~#

[Gordon Mohr]

That's atypical – indicating a memory access violation of the type that shouldn't happen in pure python, and definitely not on a mere import. 

So there's not enough info to know what's uniquely wrong in your setup, and the best approach would be to try to ensure your setup is as much like more typical setups as possible. I'd suggest:

- uninstall all (or start fresh where nothing was installed)

- install the minimum necessary (I prefer miniconda to anaconda for this reason)

- use a conda environment as a regular user– rather than installing things globally or working as root

If there's still a problem, starting python with `-v` may provide more information about what was happening just before the segmentation fault. 

- Gordon

[Jason Kim]

Hi, Gordon

I've tried as you recommend.

It's perfectly work. Thank you so much.

Below is job history.

>> rm -rf /root/anaconda2

>> cd ./download

>> bash Anaconda2-4.1.1-Linux-x86_64.sh

>> conda create -n mskim python=2.7

>> source activate mskim

>> conda install -n mskim gensim

>> python -c "import gensim"

2016년 8월 9일 화요일 오전 7시 38분 21초 UTC+9, Gordon Mohr 님의 말:

[Melvyn Drag]

I have the same issue. I had anaconda in my /opt so that it was available to all users and had this issue. 

I am only able to get this work when anaconda is installed for a user as you suggested.

 This is undesirable for the cluster where all users are depending on this python install.

Im staying on the look out and experimenting to try and resolve this issue