Intel MKL FATAL ERROR: Cannot load libmkl_def.so

[xiaoch2004]

I got this when I was trying to run the example sample_work.yaml at localhost:8888 The worker and master are both started successfully. But when I run

python kaldigstserver/client.py -r 32000 test/data/english_test.raw

I got the "Cannot load libmkl_def.so" error. I'm sure all mkl environmental paths are set to the directory which contains libmkl_def.so (/opt/intel/mkl/lib/intel64 in my case)

There are some answers from https://github.com/BVLC/caffe/issues/3884 but they are using anaconda which was not my case. I used native python 2.7.12 under ubuntu 16.04 LTS.

I thought of two potential ways to solve this. Update the python package that might call mkl or Locate where is the program looking for libmkl_def.so and copy it there Would anyone provide some help on how to do this?

[RaniRaven]

This is easy to solve. Look for this error in google, and you'll find the settings as : export LD_PRELOAD="" do it, and it'll be fine. It is some problem of the intel MKL library.

[ghwn]

In my case export LD_PRELOAD=/opt/intel/mkl/lib/intel64/libmkl_core.so:/opt/intel/mkl/lib/intel64/libmkl_sequential.so helped for me.