Open muziejus opened 7 years ago
We're currently considering switching to Anaconda for the default Python installs. NLTK is part of that package, as is Scipy and Numpy. Let us know if you have any opinions on that, and if that would resolve your issue.
I imagine it would. Is there a ballpark on the deployment of the switch? (as in, a few weeks or much longer than that?)
I'm working on it now. I'd say a few weeks would be a fair estimate.
DH Box has been updated to use Anaconda for its Python install. To update the DH Box seed, go to the /dhbox
directory and type python manage.py start
.
I did the above, with a (finally) fully operational dhbox install, but I still had to install nltk separately so that python3 in the Jupyter notebook would see it.
We've identified something going on with Docker such that it is failing to download the newest DH Box image, giving the error: layers from manifest don't match image configuration
. I'm looking into it and I'll let you know when I've got it working again.
OK, it should be possible to download the the DH Box seed again.
I updated the seed today, but I still have no ntlk when I use the jupyter notebook. As in, it's not built in.
The demo box on dhbox.org also doesn't have ntlk with the jupyter notebook.
Hm, I didn't realize that the Anaconda install of Python needs to be activated each time you log in. Currently if you go to the command line and type source activate root
, you'll be using the new Python installation, and find that NLTK is already available. I'll look into making activation unnecessary.
OK, I've updated the seed with a line that activates the correct Python version. Thanks for catching that.
It's possible to install nltk for Python 3 from the command line w/ pip3 (or whatever), but since the Jupyter notebook defaults to Python 3, it seems like the DHBox should come with nltk pre-installed for Python 3 as well.