A implementation of PURS with contextual data.

Dependency: if cpu only: pip install tensorflow==1.14

if use gpu: pip install tensorflow-gpu==1.14 conda install cudatoolkit

example command line:

cd /path/to/project/code

python train_purs.py --device /gpu:4 --batch-size 1000 --dataset jester

device = {/cpu:0, /gpu:x}

dataset = {beer, beer_small, beer_ultrasmall, jester, jester_small}