Because of the term `2.0**THRESHOLD`, 0.0 `THRESHOLD` values are currently impossible

DavidUdell / sparse_circuit_discovery

Circuit discovery in GPT-2 small, using sparse autoencoding

MIT License

6 stars 1 forks source link

Because of the term `2.0**THRESHOLD`, 0.0 `THRESHOLD` values are currently impossible #103

Closed DavidUdell closed 2 months ago

DavidUdell commented 2 months ago

I think this is a good UI choice, actually. I should just restore the ability to have precise 0.0 threshold values by leaving THRESHOLD=None. And maybe THRESHOLD should now be called THRESHOLD_EXP.