running a python script where dependencies are not avail: distributed computing

I have access to a grid (running condor) that would (potentially) allow to very substantially reduce how long by nltk based nlp tasks take. unfortunately, i dont have root access on 开发者_运维技巧the cluster so cannot install new packages, only run whatever is available on the linux boxes.

python is of course available, but nltk isnt - i was wondering however, if there might be a way around this somehow ? is there a way i can somehow still distribute the task in a self-contained 'package' of some sort?

Thanks for your hel

If you can get a standard user account, you could use virtualenv to create a sandbox in that user-account, where you can install nltk.