quantlab.model.training_target#
The training target, computed once per bar before a fit: shared by both model variants.
TrainingTargetMixin gives TorchModel and LibraryModel the one _transform_target
hook and the code around it: the collected panel as a TrainingPanel of torch tensors, and
each bar’s raw labels turned into the training target and its mask. It lives in the model
layer, not on BaseModel, so importing the root class (quantlab.model.base)
never imports torch.
Classes
The |