microsoft/qlib · error · AttributeError

DataHandlerLP has not attribute _data, please set drop_raw =

Error message

DataHandlerLP has not attribute _data, please set drop_raw = False if you want to use raw data

What it means

DataHandlerLP with `drop_raw=True` deletes the original `_data` after processing to save memory. Later access to the raw key (`data_key=DataHandler.DK_R`, e.g. handler.fetch(..., data_key=DK_R) or get_infer_data-adjacent raw reads) raises AttributeError telling you to keep the raw data if you need it.

Source

Thrown at qlib/data/dataset/handler.py:667

        # init raw data
        super().setup_data(**kwargs)

        with TimeInspector.logt("fit & process data"):
            if init_type == DataHandlerLP.IT_FIT_IND:
                self.fit()
                self.process_data()
            elif init_type == DataHandlerLP.IT_LS:
                self.process_data()
            elif init_type == DataHandlerLP.IT_FIT_SEQ:
                self.fit_process_data()
            else:
                raise NotImplementedError(f"This type of input is not supported")

        # TODO: Be able to cache handler data. Save the memory for data processing

    def _get_df_by_key(self, data_key: DATA_KEY_TYPE = DataHandlerABC.DK_I) -> pd.DataFrame:
        if data_key == self.DK_R and self.drop_raw:
            raise AttributeError(
                "DataHandlerLP has not attribute _data, please set drop_raw = False if you want to use raw data"
            )
        df = getattr(self, self.ATTR_MAP[data_key])
        return df

    def fetch(
        self,
        selector: Union[pd.Timestamp, slice, str] = slice(None, None),
        level: Union[str, int] = "datetime",
        col_set=DataHandler.CS_ALL,
        data_key: DATA_KEY_TYPE = DataHandler.DK_I,
        squeeze: bool = False,
        proc_func: Callable = None,
    ) -> pd.DataFrame:
        """
        fetch data from underlying data source

        Parameters

View on GitHub (pinned to 79633dd950)

Solutions

Set `drop_raw=False` in the handler config.
Or switch the request to DK_I (infer) / DK_L (learn) if raw is truly unneeded.

Example fix

# before
handler = DataHandlerLP(..., drop_raw=True)
df = handler.fetch(data_key=DataHandler.DK_R)

# after
handler = DataHandlerLP(..., drop_raw=False)
df = handler.fetch(data_key=DataHandler.DK_R)

Defensive patterns

Strategy: validation

Validate before calling

from qlib.data.dataset.handler import DataHandler

def raw_available(handler) -> bool:
    return not getattr(handler, 'drop_raw', False) or hasattr(handler, '_data')

Type guard

from qlib.data.dataset.handler import DataHandler

def can_fetch_raw(handler) -> bool:
    return not getattr(handler, 'drop_raw', False)

Prevention

Only enable drop_raw when nothing downstream needs DK_R.
Grep your pipeline for data_key=DK_R before turning on drop_raw.

When it happens

Trigger: Constructing a handler with `drop_raw: true` in config, then calling `handler.fetch(..., data_key=DataHandler.DK_R)` or anything requesting self.DK_R.

Common situations: Memory-tuned production configs that add drop_raw, while downstream code (e.g. some benchmarks or custom processors) still asks for raw data; also forgetting drop_raw was set in a base config.

Related errors

AI-assisted analysis of microsoft/qlib@79633dd950 (2026-08-15). Data as JSON: /api/errors/3230e1f138d2f55b. Report an issue: GitHub.