[NLPL Task Force (A)] Tensorflow issues, pt. 2
Stephan Oepen
oe at ifi.uio.no
Mon Oct 28 11:37:19 UTC 2019
> $ pip3 install --user -r requirements.txt
> ERROR: Can not perform a '--user' install. User site-packages are not visible in this virtualenv.
i would recommend you make it a habit to say 'python3 -m pip' instead,
as the 'pip3' convenience script appears to not always be installed
with new virtual environments (which is what the NLPL modules are).
this is a bit hard for me to test, because i do have write access to
the NLPL modules, but please try the following:
$ module purge; module load Python/3.7.2-GCCcore-8.2.0
$ type -path python3
/cluster/software/Python/3.7.2-GCCcore-8.2.0/bin/python3
$ python3 -m virtualenv vinitr
$ . vinitr/bin/activate
$ python3 -m pip install telegram
$ ls -ld vinitr/lib/python3.7/site-packages/telegram*
drwxrwxr-x 3 oe oe 2 Oct 28 12:31 vinitr/lib/python3.7/site-packages/telegram
drwxrwxr-x 2 oe oe 5 Oct 28 12:31
vinitr/lib/python3.7/site-packages/telegram-0.0.1.dist-info
$ module load nlpl-tensorflow/1.15.0/3.7 nlpl-horovod/0.18.2/3.7
$ . vinitr/bin/activate
$ python3 -m pip freeze | egrep 'horovod|telegram'
horovod==0.18.2
telegram==0.0.1
it will be important to get the right 'python3' binary at each point
in time, i.e. at first the system-wide one (but using the exact same
version as the NLPL modules are built from), later your own copy from
your virtual environment. use 'type --path' to confirm you are indeed
getting the right binaries!
oe
More information about the infrastructure
mailing list