Dipy install on computer cluster

1

Hi All,

for those who work with dipy:

how can dipy be installed on a linux computer cluster where there’s limited or no access to “pip” or other standard OS installation methods ? Is there something like a tar.gz file?

Best regards,
Alex

2

This resource might help! Installing from source for Unix http://nipy.org/dipy/installation.html#install-from-source-for-unix-e-g-linux-osx

You could probably make a local installation for your account on your cluster using the above instructions.

3

I generally make a “local” installation of anaconda for the group in our partition. But you could also install anaconda for your user only. Then you will be able to conda install dipy (probably after enabling the conda-forge).

4

Hi oestaban, that’s a good idea. Indeed, when Anaconda is installed already by the cluster admin - that might work. Otherwise - I can’t install anaconda - no sudo access.

5

Hi nooralsh,

thanks for the idea. I tried that but got two problems - when “cloning” - the cluster has to create a “git” folder, and it’s not allowed.

When downloading the zip from github and initiating using “python setup.py” - can’t install everything, since it needs access to /usr/lib, and I don’t have that access. When asking this question - I’m thinking about fsl or freesurfer, which can be easily installed without sudo.

6

I have successfully installed anaconda without elevated privileges. When you install, make sure you point the installation to a path you can write to. After that, everything should happen at user level.

In particular:

bash /path/to/Anaconda3-5.1.0-Linux-x86_64.sh -p /path/to/owned/folder/

Cheers

7

Hi oesteban, thanks a lot for your reply. It worked !!
There are still some challenges: I pointed to one path (e.g.: /working-folder/folder/) but it installed to a folder in the home directory (/home/user/anaconda3). Nevertheless, dipy was installed and imported. Hopefully it will work. There will probably be some challenges with the qsub command when submitting to the cluster pipeline, but at least the first stage was done ! Thanks a lot !