修复拼音组处理逻辑，避免未处理拼音导致的索引错误

src/suinput/dataset.py

@@ -440,7 +440,9 @@ class PinyinInputDataset(IterableDataset):
                 "char_id": torch.tensor([char_info["id"]]),
                 "char": char,
                 "freq": char_info["freq"],
-                "pg": torch.tensor([self.pg_groups[char_info["pinyin"][0]]]),
+                "pg": torch.tensor(
+                    [self.pg_groups[processed_pinyin[0]] if processed_pinyin else 8]
+                ),
             }
 
             # 根据调整因子重复样本