This repo contains the lecture materials for "Stanford CS336: Language modeling from scratch".

Located in nonexecutable/as PDFs

Located as lecture_*.py in the root directory

You can compile a lecture by running:

    python execute.py -m lecture_01

which generates a var/traces/lecture_01.json and caches any images as appropriate.

However, if you want to run it on the cluster, you can do:

    ./remote_execute.sh lecture_01

which copies the files to our slurm cluster, runs it there, and copies the results back. You have to setup the appropriate environment and tweak some configs to make this work (these instructions are not complete).

If you need to tweak the Javascript:

Install (one-time):

    npm create vite@latest trace-viewer -- --template react
    cd trace-viewer
    npm install

Load a local server to view at http://localhost:5173?trace=var/traces/sample.json:

    npm run dev

Deploy to the main website:

    cd trace-viewer
    npm run build
    git add dist/assets
    # then commit to the repo and it should show up on the website