- Install the public version of rust_circuit from the GitHub repository and follow the setup instructions there.
- Install additional packages for remix with
python -m pip install -r requirements.txt
- Download the models and data (about 700MB compressed) from this link and unzip to your home directory such that you have a
~/tensors_by_hash_cache/
directory.
- W0D1 - pre-course exercises on PyTorch and einops (CPU)
- D1 - Intro to Circuits (CPU)
- D2 - Build your own GPT in Circuits (CPU)
- D3 - Induction Heads Investigation (GPU preferred)
- D4a - Intro to Causal Scrubbing (GPU preferred)
- D4b - Paren Balencing (GPU preferred)
- D5 - Indirect Object Identification and Causal Scrubbing (GPU preferred)