[Tutor] Moving a conda environment to an off-line computer

[Oscar Benjamin]

On Sat, 1 Dec 2018 at 00:10, Henrique Castro <henriquecsj at outlook.com> wrote:
>
> Dear colleagues,
> Soon I'll start to use one of the powerful computers on my university as a tool in my Ph.D. The computer does not have an internet connection and I need to find a way to install a conda environment on it.

When you say that you need to install a conda environment I imagine
that what you mean is you need to install some Python packages and you
think the easiest way is to install a conda environment. Is that
correct?

Having used a similar setup at my University the situation we have is
that there is a big cluster with thousands of nodes that don't have
internet access but there is internet access on certain nodes called
the "login nodes" which you can ssh into. They have a filesystem that
is shared with all other nodes which means you can use git, pip etc to
get your code set up and working before submitting a job to be run on
the main cluster. Is that not the case for the setup you're using?

Also the setup we have actually provides many Python versions and
packages: I just have to activate them with a command that's something
like "module activate python-3.6 python-3.6-numpy ...".

The other aspect of our setup is that it is well equipped with
compilers, scientific libraries etc. so that it's usually straight
forward to compile e.g. numpy from source with "python setup.py
install". I don't know if I could also just transfer manylinux wheels
in there to avoid compiling as well...

So there are potentially a number of ways of achieving what you want.
With the information you've provided so far, I'm not clear what the
best way forward is. The first question is why is it that you want a
conda environment?

--
Oscar