Rate this Page
★ ★ ★ ★ ★

TensorDictBase.get_item_shape#

TensorDictBase.get_item_shape(key: NestedKey)#

Returns the shape of the entry at key, without reading the entry when possible.

For a lazy stack, a dimension whose size differs between the stacked entries has size -1.

Parameters:

key (NestedKey) – the key of the entry.

Raises:

KeyError – if the tensordict has no entry at key.

Examples

>>> import torch
>>> from tensordict import TensorDict, lazy_stack
>>> td = lazy_stack([TensorDict(x=torch.zeros(2)), TensorDict(x=torch.zeros(3))])
>>> td.get_item_shape("x")
torch.Size([2, -1])

Every tensordict class and every tensorclass has this method. For the other methods, see Methods by task.