This reviews my process of setting up tensorflow virtualenv in jupyter notebook in deepthought.ics.uci.edu Ubuntu machine for shoeprint classification project. Before this, I have successfully run the jupyter notebook script on my own MacOS and was able to train the model, but the process is pretty slow: it takes 20 min to train 15 echos with a training set of 60 images and a validation set of 40 images.
1. installation on Mac:
install virtualenv
install tensorflow with virtualenv
turn on jupyter notebook in tensorflow virtualenv (takes some uninstall first and install later, but cannot quite remember)
2. usage on Mac:
source activate tensorflow
jupyter notebook
->open the script
3. install on deepthought：
pip install virtualenv --user (this is important, which allows me to install under my own directory without having sudo authorization)
install tensorflow with virtualenv: virtualenv --system-site-packages -p python3 (may not be exact, something like this)
then install jupyter notebook:
4. usage on deepthought:
cd ~ (to my home directory)
source tensorflow/bin/activate tensorflow
5. set up jupyter notebook server and run on remote:
1. turn on jupyter notebook server:
from MacOS:
login: ssh leix8@deepthought.ics.uci.edu
activate: virtual environment: source tensorflow/bin/activate tensorflow
turn on kernel: ipython notebook --no-browser --port=8889
2. then set up a ssh channel:
from MacOS:
ssh leix8@deepthought.ics.uci.edu -L127.0.0.1:1234:127.0.0.1:8889
(then enter a password)
go to browser and enter: http://127.0.0.1:1234 or http://localhost:1234
