diff --git a/unsloth/kernels/cross_entropy_loss.py b/unsloth/kernels/cross_entropy_loss.py index 66d5046deb..9f16e8e605 100644 --- a/unsloth/kernels/cross_entropy_loss.py +++ b/unsloth/kernels/cross_entropy_loss.py @@ -273,8 +273,8 @@ class Fast_CrossEntropyLoss(torch.autograd.Function): n_chunks = div + (mod != 0) losses = torch.empty(n_rows, dtype = torch.float32, device = logits.device) - DO_SOFTCAPPING = (logit_softcapping != 0) - DO_LOGIT_SCALING = (logit_scaling != 0) + DO_SOFTCAPPING = logit_softcapping != 0 + DO_LOGIT_SCALING = logit_scaling != 0 if n_chunks == 1: # For small vocabs <= 65336 like Llama, Mistral