search

dilyabareeva / quanda

A toolkit for quantitative evaluation of data attribution methods.
https://quanda.readthedocs.io
MIT License
33 stars 0 forks source link

Remove state-methods of benchmarks #144

Closed dilyabareeva closed 2 months ago

dilyabareeva commented 2 months ago

Since there is not sensible way to save large portions of the benchmark state, we should not even offer these methods.

IMPORTANT: We still have download (name: str)method for loading actual benchmarks from our cloud / HuggingFace