PyPI page
Home page
Author:
Hem Gosalia, Aditya Kasliwal, Vinay Kumar Sankarapu
License:
LSAL-1.2
Summary:
Unified Discover, Evaluate, Intervene toolkit for mechanistic interpretability: 13 circuit discovery algorithms, 6-pillar faithfulness evaluation, and pruning / quantization / knowledge editing / unlearning / activation steering / contrastive weight steering interventions behind one API.
Latest version:
0.2.0
Required dependencies:
accelerate
|
bert-score
|
click
|
einops
|
evaluate
|
huggingface-hub
|
ipywidgets
|
matplotlib
|
networkx
|
numpy
|
ordered-set
|
pandas
|
plotly
|
psutil
|
pyyaml
|
rich
|
rouge-score
|
sacrebleu
|
safetensors
|
scikit-learn
|
scipy
|
seaborn
|
sentence-transformers
|
spacy
|
torch
|
torch-pruning
|
tqdm
|
transformer-lens
|
transformers
Optional dependencies:
black
|
captum
|
datasets
|
flake8
|
isort
|
lime
|
lm-eval
|
mkdocs
|
mkdocs-jupyter
|
mkdocs-material
|
mkdocstrings
|
mypy
|
optimum-quanto
|
pre-commit
|
pygments
|
pymdown-extensions
|
pytest
|
pytest-cov
|
pytest-timeout
|
pytest-xdist
|
shap
|
torch
|
torchvision
Downloads last day:
14
Downloads last week:
86
Downloads last month:
1,093
7-day smoothing