First things first for all setup methods is to clone this repo:
git clone https://github.com/chrishamblin7/circuit_pruner.git
Create a new conda environment from the environment_file: conda env create -f environment/environment.yml
Activate the new "circuit_pruner" environment: conda activate circuit_pruner
Add the circuit_pruner package itself to the "circuit_pruner" environment (run inside the root folder, with the setup.py file): pip install -e .
Then, if you want to use the jupyter notebooks, add the viscnn environment as a kernel with: python -m ipykernel install --user --name=circuit_pruner
Read the 'notes.md' file in the 'method_comparison' folder. For a quick look at the data run 'mrthod_comparison/plotting.ipynb' and choose the 'quick start' options.
for polysemantic experiments: 'polysemantic_subfeatures.ipynb' for circle-scale experiment: 'circle_subfeatures.ipynb'
'circuit_diagram.ipynb'
Use the quick start option to quickly acces the GUI. Will be accessible in a new browser window at 'http://localhost:8050/'