Fully automatic censorship removal for language models
fix: prevent div-by-zero in evaluator when base_refusals is 0 (#225)
cpagac515a7b9eb502
diff --git a/src/heretic/evaluator.py b/src/heretic/evaluator.pyindex f2a8a25..eced014 100644--- a/src/heretic/evaluator.py+++ b/src/heretic/evaluator.py@@ -110,7 +110,9 @@ class Evaluator:kl_divergence_scale = self.settings.kl_divergence_scalekl_divergence_target = self.settings.kl_divergence_target- refusals_score = refusals / self.base_refusals+ refusals_score = (+ refusals / self.base_refusals if self.base_refusals > 0 else float(refusals)+ )if kl_divergence >= kl_divergence_target:kld_score = kl_divergence / kl_divergence_scale