topic Re:Intel Python pip check fails on fresh install (sdc numba requirement) in Intel® Distribution for Python*

Intel Python pip check fails on fresh install (sdc numba requirement)

user02 — Thu, 04 Feb 2021 14:47:28 GMT

A fresh install of the offline, single component "Intel Distribution for Python for Linux" installer (l_pythoni37_oneapi_p_2021.1.1.44_offline.sh) creates a Python environment that fails "pip check."

> ./l_pythoni37_oneapi_p_2021.1.1.44_offline.sh -a \
--eula accept \
--install-dir ${wherever} \
--silent

> source setvars.sh
:: initializing oneAPI environment ...
BASH version = 4.3.48(1)-release
:: dal -- latest
:: mkl -- latest
:: intelpython -- latest
:: compiler -- latest
:: ipp -- latest
:: mpi -- latest
:: tbb -- latest
:: oneAPI environment initialized ::

> python -m pip check
sdc 0.37.0 has requirement numba<0.52,>=0.51.2, but you have numba 0.51.2-dppy0.10.1rc1-0.gb5af1f5d9.dirty.

Installer distributed by Intel from: https://software.intel.com/content/www/us/en/develop/articles/oneapi-standalone-components.html#python

Presumably, this is just an artifact of how pip evaluates the dppy dirty build of numba 0.51.2. However, it would be best if a fresh install passed "pip check" with no warnings or errors.

Re:Intel Python pip check fails on fresh install (sdc numba requirement)

AthiraM_Intel — Fri, 05 Feb 2021 13:59:50 GMT

Hi,

Thanks for posting in Intel Distribution for python forum and reporting this issue. We will inform the concerned team regarding this. We will keep you posted on updates.

Thanks.

Re:Intel Python pip check fails on fresh install (sdc numba requirement)

Rachel_O_Intel — Tue, 08 Mar 2022 16:04:01 GMT

Hi,

I am here to provide an update to this case.

It turns out that "pip check" works sometimes incorrectly for conda environment.

Generally, pip cannot be truly used for checking conda environment. The IDP environment was created based on conda environment and if the environment exists - then all packages there are compatible.

Conda ensures about it during environment creation. That is why we cannot recommend to use "pip check" for checking conda environment.

At the same time we are going to investigate if it is possible to improve this check to make it pass.

Please check back in future updates for Intel Distribution of Python for improvements.

This issue is now closed. If you need any additional information, please post a new question as this thread will no longer be monitored by Intel.

Best,

Rachel