feat : 数据流水拼接策略支持 position_ids 预计算

- OutputConfig.position_ids_mode 三种模式控制边界策略
- pipeline._flush() 按配置生成扁平 position_ids 数组
- SFTDataset 在 __getitem__ 中返回 position_ids
- SFTStrategy 将 position_ids 传入 model.forward()

astrai/config/preprocess_config.py

@@ -45,6 +45,13 @@ class OutputConfig(BaseConfig):
     storage_format: str = "bin"
     max_tokens_per_shard: int = 100_000_000
     dtype: Dict[str, str] = field(default_factory=dict)
+    position_ids_mode: Optional[str] = None
+    """How to compute position_ids in packed sequences.
+
+    - ``None`` / ``"none"``: do not generate (backward compatible).
+    - ``"doc_reset"``: reset to 0 at each document boundary.
+    - ``"continuous"``: sequential 0, 1, 2, ... (pretrain, single doc).
+    """
 
 
 @dataclass