This PR includes two small changes to improve the regression interface.
First, I added a default value for context_size, the same as for classification.
Second, I fixed an error in the case where the context was larger than the training data. Previously, this would cause an error unsupported operand type(s) for *: 'numpy.ndarray' and 'Tensor' as the predictions were transformed to numpy before being multiplied with a torch tensor.
Heyho,
This PR includes two small changes to improve the regression interface.
First, I added a default value for
context_size, the same as for classification.Second, I fixed an error in the case where the context was larger than the training data. Previously, this would cause an error
unsupported operand type(s) for *: 'numpy.ndarray' and 'Tensor'as the predictions were transformed to numpy before being multiplied with a torch tensor.