Frequently Asked Questions
When is the program running next / how can I apply?
For this and related questions, see our informational site, which contains more information about the program, when we're opening applications, who is on our team, and more.
Does this page replace the Streamlit page?
You can use either this website or the Streamlit page - both will still exist & be supported going forwards (they both pull material from the same GitHub repo, which is also the repo that hosts the exercises).
Going forward, we recommend people use this website rather than the Streamlit page, because it provides an improved user experience (e.g. better chatbot, ability to download exercise context, etc). We created it because the course was getting too large & complex for Streamlit to give us the flexibility we wanted.
How to study this material?
You can study this material in the exact same way as before. There are 3 options:
- Use Colabs - each chapter section will provide a link to an "exercise Colab" (for filling in the answers to questions) and a "solutions Colab" (where you can compare your solutions to the ones we've provided).
- Use VSCode (or your own IDE), working from empty Python files - you can read through the material on this website, and copy in blocks of code to the notebook or Python file you're working with. This will require cloning the ARENA_3.0 repo (and we also recommend using a remote machine for some of the more compute-intensive exercise days).
- Use VSCode (or your own IDE), working from the exercise notebooks - this is essentially a hybrid of the first 2 approaches: you can work on your preferred IDE, but complete the exercises by working through
.ipynbexercise files which are equivalent to the "exercise Colab" notebooks mentioned above. Again, this will require cloning the ARENA_3.0 repo (and we recommend using a remote machine). Each exercise notebook can be found in a path likechapter1_transformer_interp/exercises/part1_transformer_from_scratch/1.1_Transformer_from_Scratch_exercises.ipynb
For detailed setup instructions (including how to set up a virtual machine), see the Setup Instructions page.
How can I contribute to this website?
This website is open source! If you find bugs, want to suggest improvements, or would like to contribute code, visit the arena-heroku GitHub repository.
- Report issues - open an issue on GitHub describing the bug or suggestion.
- Submit a pull request - fork the repo, make your changes, and open a PR. We welcome contributions of all kinds: bug fixes, UI improvements, new features, and documentation.
- Improve the exercises - the exercise content itself lives in the ARENA_3.0 repo. If you spot errors or want to improve exercise material, contribute there instead.