I upgraded Numpy (1.16.6) and Pandas(1.1.5).
When I run python with django + apache + mod_wsgi , it fails with error: Intel MKL FATAL ERROR: Cannot load <mkl-loader>.
I've installed MKL( l_onemkl_p_2021.1.1.52_offline.sh ).
Red Hat Enterprise Linux Server release 7.4 (Maipo)
Intel(R) Xeon(R) CPU E5-2660 v3 @ 2.60GHz
I did not find actual source line containing <mkl-loader>, and my python project log is empty.
It seems that python is loading some lib, but the error does not show which lib loaded fail, maybe it is numpy.
I rebuild numpy and scipy by intel document, as follow, but it not fix error.
This is apache error log.
How can I fix this error? Thanks.
Thanks for reaching out to us.
Could you please install Intel Distribution for Python and use the MKL along with the python distribution for running the program.
You can follow the below link to install Intel Distribution for Python:
Please uninstall the MKL you used, before installing the Intel Distribution for Python.
If the issue still persists please let us know.
I have taken your advice, but it did not fix the error.
After I fixed another error, as follows, the error "Cannot load <mkl-loader>" occurred again.
Is that apache or mod_wsgi not supported mkl?
Or where should I post this question?
Could you please confirm whether you used a fresh conda environment?
Please verify that the python path used is correct using the below commands:
Make sure that all the installation steps are successfully completed.
Also please share the conda and pip list.
We will check with the Subject Matter Experts about your issue and will post the updates.
I use Anaconda3 (Anaconda3-5.2.0-Linux-x86_64.sh ).
conda version is 4.5.4
And I create a virtul env by command : python -m venv xxx.
This is python path information :
The pip list in virtual env is as follows:
From the screenshots we could see that you have mixed environment , multiple mkls in the python (3.6.5).
Also we can see that you are not using Intel Distribution for Python.
Could you please change your environment to intel python (3.7), then check if the issue is reproduced.
If the issue still persists, please share the reproducer instructions.