microsoft/qlib · error · ValueError

LightGBM doesn't support multi-label training

Error message

LightGBM doesn't support multi-label training

What it means

Thrown by HFMLGBModel._prepare_data when the label block is not exactly one column. The model converts the single label into a per-timestamp alpha and then a binary label_c (0/1 by sign) for the LightGBM classifier; multiple label columns leave no unambiguous target, and LightGBM needs a 1D label anyway.

Source

Thrown at qlib/contrib/model/highfreq_gdbt_model.py:110

            # Convert label into alpha
            df_train.loc[:, ("label", l_name)] = (
                df_train.loc[:, ("label", l_name)]
                - df_train.loc[:, ("label", l_name)].groupby(level=0, group_keys=False).mean()
            )
            df_valid.loc[:, ("label", l_name)] = (
                df_valid.loc[:, ("label", l_name)]
                - df_valid.loc[:, ("label", l_name)].groupby(level=0, group_keys=False).mean()
            )

            def mapping_fn(x):
                return 0 if x < 0 else 1

            df_train["label_c"] = df_train["label"][l_name].apply(mapping_fn)
            df_valid["label_c"] = df_valid["label"][l_name].apply(mapping_fn)
            x_train, y_train = df_train["feature"], df_train["label_c"].values
            x_valid, y_valid = df_valid["feature"], df_valid["label_c"].values
        else:
            raise ValueError("LightGBM doesn't support multi-label training")

        dtrain = lgb.Dataset(x_train, label=y_train)
        dvalid = lgb.Dataset(x_valid, label=y_valid)
        return dtrain, dvalid

    def fit(
        self,
        dataset: DatasetH,
        num_boost_round=1000,
        early_stopping_rounds=50,
        verbose_eval=20,
        evals_result=None,
    ):
        if evals_result is None:
            evals_result = dict()
        dtrain, dvalid = self._prepare_data(dataset)
        early_stopping_callback = lgb.early_stopping(early_stopping_rounds)
        verbose_eval_callback = lgb.log_evaluation(period=verbose_eval)

View on GitHub (pinned to 79633dd950)

Solutions

1. Configure exactly one label expression in the data handler
2. Move any extra targets into features or pick a multi-output-capable model

Example fix

# before
label: ["Ref($close, -2)/Ref($close, -1) - 1", "$vwap/$close - 1"]

# after
label: ["Ref($close, -2)/Ref($close, -1) - 1"]

y = dataset.prepare("train", col_set="label", data_key="learn")
assert y.values.ndim == 2 and y.values.shape[1] == 1, "HFMLGBModel needs a single label column to build the binary target"

When it happens

Trigger: Fitting HFMLGBModel with a handler label list containing more than one expression, so y_train.values.shape[1] != 1.

Common situations: Reusing a multi-label handler config from another model with the high-frequency workflow; adding auxiliary label columns for analysis.

Related errors

• LightGBM doesn't support multi-label training
• LightGBM doesn't support multi-label training
• Model hasn't been trained yet
• CatBoost doesn't support multi-label training
• Empty data from dataset, please check your dataset config.

AI-assisted analysis of microsoft/qlib@79633dd950 (2026-08-15). Data as JSON: /api/errors/6be0c4ad8ca75a7c. Report an issue: GitHub.