TensorDictBase.densify#
- TensorDictBase.densify(layout: layout = torch.strided)#
Returns the tensordict with its lazy stacks replaced by dense tensordicts.
The entries of a lazy stack are stacked into one tensor when their shapes match, and into a nested tensor (
torch.nested) along the stack dimension otherwise. A key that only some members of a lazy stack have is filled with an empty entry for the members that lack it (zero rows when the members that have it agree on its feature dims), so it appears in the result althoughkeys()of the lazy stack leaves it out. A tensordict that holds no lazy stack is returned as it is.- Parameters:
layout (layout, optional) – the layout of the nested tensors,
torch.stridedortorch.jagged. Defaults totorch.strided.
Examples
>>> import torch >>> from tensordict import TensorDict >>> stack = TensorDict.lazy_stack( ... [TensorDict(a=torch.zeros(3)), TensorDict(a=torch.ones(4))] ... ) >>> dense = stack.densify(layout=torch.jagged) >>> type(dense).__name__, dense["a"].is_nested ('TensorDict', True)
Every tensordict class and every tensorclass has this method. For the other methods, see Methods by task.
