Simple but flexible Deep Recommenders in PyTorch. This hands-on tutorial will teach you:
Check out the notebooks within to step through variations of matrix factorization models.
If at all possible, please check out and pre-install the environment.
git clone https://github.com/cemoody/simple_mf.git cd simple_mf
Create the environment by following the steps below. If you choose to use your own environment, you’ll need access to have the Python packages in
Follow the directions the above command spits out.
The output of the last command, which depends on if you’re using
conda or not, will tell you how to activate your environment.
Follow the steps in #1 carefully – you probably don’t need step 2a. or 2b!
If the above step used conda, you can active it the conda environment by running:
source activate simple_mf
If you don’t have
conda, then the output of the
make create_environment will spit out something like this:
Make sure the following lines are in shell startup file export WORKON_HOME=/Users/chrismoody/.virtualenvs export PROJECT_HOME=/Users/chrismoody/Devel source /usr/local/bin/virtualenvwrapper.sh workon simple_mf
Go ahead and activate your environment by running the above commands.
This will download and preprocess the MovieLens 1M dataset. We’ll use this canonical dataset to test drive our code.
While we’re using PyTorch instead of Tensorflow directly, the logging and visualization library Tensorboard is an amazing asset to track the progress of our models. It’s implemented as a small local web server that constructs visualizations from log files, so start by kicking it off in the background:
cd notebooks tensorboard --logdir runs
Visit the tensorboard dashbaord by going to http://localhost:6006
This will startup the Jupyter server and open up the available notebooks. Try running a few notebooks ahead of time to verify that your environment is setup and functioning well.
Visit the jupyter notebooks by going http://localhost:8888/tree