HW 7: Neural Networks for CS349 @ NU

IMPORTANT: PUT YOUR NETID IN THE FILE netid in the root directory of the assignment. This is used to put the autograder output into Canvas. Please don't put someone else's netid here, we will check.

In this assignment, you will:

• Learn how to use PyTorch
• Implement two types of Neural Networks
• Explore the impact of different hyperparameters on your networks' performance

Clone this repository

To clone this repository install GIT on your computer and copy the link of the repository (find above at "Clone or Download") and enter in the command line:

git clone YOUR-LINK

Alternatively, just look at the link in your address bar if you're viewing this README in your submission repository in a browser. Once cloned, cd into the cloned repository. Every assignment has some files that you edit to complete it.

Files you edit

See problems.md for what files you will edit.

Do not edit anything in the tests directory. Files can be added to tests but files that exist already cannot be edited. Modifications to tests will be checked for.

Jupyter Notebook

We have created a jupyter notebook which you can use that should guide you through the complete homework. We heavily recommend using it, since most Deep Learning research is now done within Jupyter Notebooks. You find the notebook here:

experiments/deep-learning-experiments.ipynb

Environment setup

Make a conda environment for this assignment, and then run:

pip install -r requirements.txt

IMPORTANT: PyTorch is not included in requirements.txt! To install PyTorch, find the correct install command for your operating system and version of python here. For "PyTorch Build" select the Stable (1.1) build, select your operating system, for "Package" select pip , for "Language" select your python version (either python 3.5, 3.6, or 3.7), and finally for "CUDA" select None. Make sure to run the command with your conda environment activated.

Running the test cases

The test cases can be run with:

python -m pytest -s

at the root directory of the assignment repository.

Questions? Problems? Issues?

Simply open an issue on the starter code repository for this assignment here. Someone from the teaching staff will get back to you through there!