Model Checkpoints + Randomization

dilyabareeva / quanda

A toolkit for quantitative evaluation of data attribution methods.

https://quanda.readthedocs.io

MIT License

33 stars 0 forks source link

Open dilyabareeva opened 2 months ago

dilyabareeva commented 2 months ago

Some explainers load from checkpoints, so randomizing the model attribute does not make sense.

dilyabareeva commented 1 month ago

Possibly combinable with #18.