fix #4820

src/llamafactory/train/sft/metric.py

@@ -55,7 +55,15 @@ def compute_accuracy(eval_preds: "EvalPrediction") -> Dict[str, float]:
 
 
 def eval_logit_processor(logits: "torch.Tensor", labels: "torch.Tensor") -> "torch.Tensor":
-    logits = logits[0] if isinstance(logits, (list, tuple)) else logits
+    if isinstance(logits, (list, tuple)):
+        if logits[0].dim() == 3:  # (batch_size, seq_len, vocab_size)
+            logits = logits[0]
+        else:  # moe models have aux loss
+            logits = logits[1]
+
+    if logits.dim() != 3:
+        raise ValueError("Cannot process the logits.")
+
     return torch.argmax(logits, dim=-1)