Assignment 6: Deep Learning
Due: At the end of day11:59pm, Tuesday, April 7, 2020.
The purpose of this assignment is to introduce you to deep learning. Specifically, the assignment consists of three parts: In the first part, you will implement various PyTorch deep learning layers using Numpy; in part two, you will experiment with different hyper-parameters on a image classification task and find the best hyper-parameters; lastly, you will investigate a state-of-the-art neural architecture from the PyTorch model zoo.
The assignment
Since we will be experimenting with some cutting edge models, to avoid your need to install dependencies, we will be using Google Colab. Google Colab is a free online computing platform that provides free GPU access to run Python code. The format of Colab is based on Jupyter notebooks. There will be a short intruction about how to interact with Colab, which should probably take around 5 minutes. Should you encountered any problems, feel free to make a post on Piazza or attend TA office hours. Our TAs are experts in PyTorch!
IMPORTANT: Your progress in Colab is not saved automatically. You can save your Colab notebook by either downloading it or by updating it in you Google Drive.
Instructions to kick start the assignment:
- Download the assignment package hw6.zip
- Unzip hw6.zip.
- Open Google Colab here (https://colab.research.google.com)
- Click "Upload" and upload the file deep_learning.ipynb.
- Click on the "folder-like" icon on the left and click "Upload"
- Add the files (hw_utils.py, birb.jpg) to colab.
Deliverables
Hand in your Jupyter Notebook. You do not need to hand in a separate PDF writeup.