It's not totally clear how this should work, but the trees are best interpreted in the context of a specific sampler, so it seems funny not to store that. (This is especially true for OutOfTimeSampler, for example, where it would be better to store the time threshold for later use).
It's not totally clear how this should work, but the trees are best interpreted in the context of a specific sampler, so it seems funny not to store that. (This is especially true for OutOfTimeSampler, for example, where it would be better to store the time threshold for later use).