TensorDictBase.detach#
- TensorDictBase.detach() Self#
Returns a new tensordict with every tensor leaf detached from the autograd graph.
The leaves are
torch.Tensor.detach()views: they share their storage with the leaves of the tensordict, and do not require gradients. A locked tensordict gives a locked result.Examples
>>> import torch >>> from tensordict import TensorDict >>> td = TensorDict(a=torch.zeros(3, requires_grad=True) + 1) >>> out = td.detach() >>> out["a"].requires_grad, td["a"].requires_grad (False, True) >>> out["a"].data_ptr() == td["a"].data_ptr() True
Every tensordict class and every tensorclass has this method. For the other methods, see Methods by task.
