AI‐24sp‐2024‐04‐11‐Afternoon - TheEvergreenStateCollege/upper-division-cs-23-24 GitHub Wiki

Goals in Class Today

Work with your partner on the assigned prototype below.
Finish typing from the Mike Nielsen notes into network.py and the load_mnist.py
Download the MNIST data from Yann Lecun's website again if you don't have it on this workspace.
Load the training data and begin training a model on your neural network.

We will continue typing and updating the MNIST Classifier from Mike Nielsen's notes. You will be assigned a random prototype from the list below:

Start your GitPod workspace, or open your laptop to your locally cloned upper-division-cs on the main branch, to a clean working directory.
Change to your assigned prototype directory and gauge how far along it is, and what you can add to it today. Read the README.md file.

cd <repo_dir>/ai-24sp/prototypes/prototype-xx

Extract the data so the files end in *ubyte and not *.gz

pip3 install numpy pillow

Run load_mnist.py and make sure that it creates one file mnist.png that you can view in VSCode.

$ python3 load_mnist.py

Finish typing network.py from the Mike Nielsen notes.
Add a function to load_mnist to load all images, reshape them as Numpy arrays, and zips them together with labels.

We will do this part on screen together in lab.

...

Add a function to import network and instantiate a neural network with the training and validation data in the expected format.
Run load_mnist again. You should see epoch training messages.

Epoch 0: 9129 / 10000
Epoch 1: 9295 / 10000
Epoch 2: 9348 / 10000
...
Epoch 27: 9528 / 10000
Epoch 28: 9542 / 10000
Epoch 29: 9534 / 10000

Stretch Goal: Save the trained model (weights and biases) as a Python3 pickled object, so that we can see its file size, and you can ship it to another team next week.