datetime occurs multiple times

[bsq1989]

🐛 Bug Description

when change example config to alpha360, datetime occurs multiple times error occurs. i debug to recognize may be record_temp.py can fix to solve

To Reproduce

Steps to reproduce the behavior:

1.change the data handler to use alpha360

    dataset:
        class: DatasetH
        module_path: qlib.data.dataset
        kwargs:
            handler:
                class: Alpha360
                module_path: qlib.contrib.data.handler
                kwargs: *data_handler_config
            segments:
                train: [2022-04-12, 2024-08-12]
                valid: [2024-08-13, 2024-12-31]
                test: [2025-01-05, 2025-04-01]

2. qrun config

Expected Behavior

no error

Screenshot

Environment

Note: User could run cd scripts && python collect_info.py all under project directory to get system information
and paste them here directly.

• Qlib version: 0.9.6
• Python version:3.10
• OS (Windows):

Additional Notes

after i change the code from workflow/record_temp.py function _generate

    def _generate(self, **kwargs):
        pred = self.load("pred.pkl")

        # replace the "<PRED>" with prediction saved before
        placeholder_value = {"<PRED>": pred}
        for k in "executor_config", "strategy_config":
            setattr(self, k, fill_placeholder(getattr(self, k), placeholder_value))
        # if the backtesting time range is not set, it will automatically extract time range from the prediction file
        dt_values = pred.index.get_level_values("datetime")
        if self.backtest_config["start_time"] is None:
            self.backtest_config["start_time"] = dt_values.min()
        if self.backtest_config["end_time"] is None:
            self.backtest_config["end_time"] = get_date_by_shift(dt_values.max(), 1)
......

to

    def _generate(self, **kwargs):
        pred = self.load("pred.pkl")

        # replace the "<PRED>" with prediction saved before
        placeholder_value = {"<PRED>": pred}
        for k in "executor_config", "strategy_config":
            setattr(self, k, fill_placeholder(getattr(self, k), placeholder_value))
        # if the backtesting time range is not set, it will automatically extract time range from the prediction file
        datetime_level = pred.index.names.index("datetime")
        dt_values = pred.index.get_level_values(datetime_level)
        if self.backtest_config["start_time"] is None:
            self.backtest_config["start_time"] = dt_values.min()
        if self.backtest_config["end_time"] is None:
            self.backtest_config["end_time"] = get_date_by_shift(dt_values.max(), 1)
....

the error disapper

[SunsetWolf]

Hi, @bsq1989
I'd like to know how to reproduce this.

[EdenRochmanSharabi]

The fix is to resolve the datetime level position via index.names.index('datetime') instead of string-based lookup. Working on a PR.