Debugging Metrics & Benchmarks - Githubissues

dilyabareeva / quanda

A toolkit for quantitative evaluation of data attribution methods.

https://quanda.readthedocs.io

MIT License

33 stars 0 forks source link

Debugging Metrics & Benchmarks #158

Closed dilyabareeva closed 2 months ago

dilyabareeva commented 2 months ago

Closes #116 (set filters where necessary, set the defaults according to papers)
remove model_id, cache_dir from base explainer signature
make benchmark evaluateargs: explainer, expl_kwargs, batch_size
some bug fixes in the metrics