FastAI Ubuntu setup on Windows (part 2)

3 min readApr 12, 2018

Continuing from part 1.

Previously we installed the Ubuntu app on windows and Nvidia CUDA drivers.

Instead of the cloning the fast ai git repo onto your servers/computer, fork the repo onto your own github account. Then remember to keep git repos synched across multiple computers if not working exclusively on one machine.

git clone https://github.com/<your-user-name>/fastai/fastai

my default shell in the ubuntu app on windows had this horrible color scheme. (ok, some might like it)

Welcome to the world of customising bash shells, many have preceded us and created awesome things. It might appear complex, it can be simple.

Simple solution for now. In the bash shell.

LS_COLORS="ow=01;36;40" && export LS_COLORS

This is temporary, will not ‘stick’ when we exit the shell and re-enter.

To make ‘sticky’ need to edit the ~/.bashrc config file

sudo vim ~/.bashrc
#add line below to end of the fileLS_COLORS="ow=01;36;40" && export LS_COLORS

exit bash and re-enter. Color changes will be permanent.

Much much more detail on editing bash configs here, here and here. It’s a rabbit hole, don’t get lost there.

— — — — — — — — — — — — — — — — — — — — — —

Back to setting up fastai on ubuntu on windows………. the fast.ai forum instructions tell us to.

conda env create -f environment-win.yml

but environment-win.yml is missing from the git repo.

The git repo install instructions tell us to

conda env update

checking conda.io documentation and this guide and looking at the contents of environment.yml, we see only environment ‘fastai’ is defined with a long list of dependencies. I’m going with the conda env update install.