diff --git a/library/custom_train_functions.py b/library/custom_train_functions.py
index 6f7737ed..0d3da3a1 100644
--- a/library/custom_train_functions.py
+++ b/library/custom_train_functions.py
@@ -845,7 +845,7 @@ def bpo_loss(loss: Tensor, ref_loss: Tensor, beta: float, lambda_: float) -> tup
     return losses.mean(dim=(1, 2, 3)), metrics
 
 
-def kto_loss(loss: Tensor, ref_loss: Tensor, kl_loss: Tensor, ref_kl_loss: Tensor, w_t=1.0, undesireable_w_t=1.0, beta=0.1):
+def kto_loss(loss: Tensor, ref_loss: Tensor, kl_loss: Tensor, ref_kl_loss: Tensor, w_t=1.0, undesirable_w_t=1.0, beta=0.1):
     """
     KTO: Model Alignment as Prospect Theoretic Optimization
     https://arxiv.org/abs/2402.01306
@@ -884,7 +884,7 @@ def kto_loss(loss: Tensor, ref_loss: Tensor, kl_loss: Tensor, ref_kl_loss: Tenso
 
     # Undesirable (rejected) samples: we want KL > reward
     if rejected_rewards.shape[0] > 0:
-        rejected_kto_losses = undesireable_w_t * (1 - F.sigmoid(beta * (KL_estimate - rejected_rewards)))
+        rejected_kto_losses = undesirable_w_t * (1 - F.sigmoid(beta * (KL_estimate - rejected_rewards)))
         losses.append(rejected_kto_losses)
 
     if losses: