At the moment, binning must be applied to the two underlying base learners in the same way. So the index vector of bl1 must be the same as the one from bl2. This is ok as long as both rely on the same feature and use binning. But, the training crashes if one uses binning and the other doesn't.
At the moment, binning must be applied to the two underlying base learners in the same way. So the index vector of bl1 must be the same as the one from bl2. This is ok as long as both rely on the same feature and use binning. But, the training crashes if one uses binning and the other doesn't.