Chengyang

DoubleEnsemble doesn't Honor sub_weights When Computing Loss Values

2

Hello, I notice in `fit()`, `loss_values` is computed based on the average prediction of trained sub-models (i.e. `pred_ensemble`): https://github.com/microsoft/qlib/blob/a87b02619aee4aff9c5aba23a66e614106050d75/qlib/contrib/model/double_ensemble.py#L90-L91 However, in `predict()`, the final output is the weighted sum of...