diff --git a/cprt/model/cprt_model.py b/cprt/model/cprt_model.py
index 9e0ee23..91a44b8 100644
--- a/cprt/model/cprt_model.py
+++ b/cprt/model/cprt_model.py
@@ -112,7 +112,7 @@ def training_step(self, batch: CprtData, batch_idx: int) -> Tensor:
         out = self(protein_ids=batch.protein, info_ids=batch.info, attention_mask=batch.info_mask, labels=batch.labels)
         loss: Tensor = out["loss"]
         self.log("loss/train_loss", loss.item(), prog_bar=True)
-        self.train_perplexity.update(out["logits"].detach(), batch.labels)
+        self.train_perplexity.update(out["logits"].detach()[:, :-1], batch.labels[:, 1:])
         if batch_idx % self.trainer.val_check_interval == 0 and batch_idx != 0:
             self.log("metrics/train_perplexity", self.train_perplexity.compute(), on_step=True)
             self.train_perplexity.reset()
@@ -129,7 +129,7 @@ def validation_step(self, batch: CprtData, batch_idx: int) -> None:
         )
         loss: Tensor = out["loss"]
         self.log("loss/val_loss", loss.item(), prog_bar=True)
-        self.val_perplexity.update(out["logits"], batch.labels)
+        self.val_perplexity.update(out["logits"][:, :-1], batch.labels[:, 1:])
         input_text = self.text_tokenizer.batch_decode(batch.info, skip_special_tokens=True)
         generated_text = self.text_tokenizer.batch_decode(torch.argmax(out["logits"], dim=-1), skip_special_tokens=True)
         self.val_rouge_scores.update(generated_text, input_text)