def __call__(self, x, inverse=False):
if inverse:
if self.mode == 'min-max':
return apply_to_tensor(x, self.norm_min_max_inv)
if self.mode == 'standardize':
return apply_to_tensor(x, self.norm_standard_inv)
else:
if self.mode == 'min-max':
return apply_to_tensor(x, self.norm_min_max)
if self.mode == 'standardize':
return apply_to_tensor(x, self.norm_standard)
def _detach_hidden(hidden: Union[torch.Tensor, Sequence, Mapping, str, bytes]):
"""Cut backpropagation graph.
Auxillary function to cut the backpropagation graph by detaching the hidden
vector.
"""
return apply_to_tensor(hidden, torch.Tensor.detach)
def _detach_hidden(hidden):
"""Cut backpropagation graph.
Auxillary function to cut the backpropagation graph by detaching the hidden
vector.
"""
return apply_to_tensor(hidden, torch.Tensor.detach)
def __call__(self, x):
return apply_to_tensor(
x, lambda y: clamp(as_tensor(y), self.floor, self.ceil))