reference/generated/tensordict.utils.LinkedList

Run in Google Colab

Colab

Download Notebook

Notebook

View on GitHub

GitHub

LinkedList

class tensordict.utils.LinkedList(*args, td: TensorDictBase | None = None, **kwargs)

Bases: list

A list of the values of a non-tensor entry that writes item assignments back to the tensordict.

td[key] returns a LinkedList when the entry is a NonTensorStack, and so does tolist() with as_linked_list=True. Setting an item (values[0] = "z") writes the value to the tensordict. The methods that change the length of the list (append, extend, insert, pop, remove, clear and del) change the list only, and unlink it from the tensordict: later item assignments no longer write back. A LinkedList compares equal to a list with the same items. If the write to the tensordict raises, the list is unchanged. tolist() returns a plain list copy that is not linked to the tensordict, as tolist() does. LinkedList(iterable), without td, builds a list that is not linked to any tensordict.

Examples

>>> from tensordict import TensorDict
>>> td = TensorDict(batch_size=[3])
>>> td["names"] = ["a", "b", "c"]
>>> values = td["names"]
>>> values[0] = "z"
>>> td.get("names").tolist()
['z', 'b', 'c']
>>> values.tolist()
['z', 'b', 'c']

append(object: Any) → None

Appends object to the list and unlinks the list from the tensordict.

clear() → None

Removes every item and unlinks the list from the tensordict.

extend(object: Any) → None

Extends the list with the items of object and unlinks it from the tensordict.

insert(index: int, object: Any) → None

Inserts object before index and unlinks the list from the tensordict.

pop(index: int = -1) → Any

Removes and returns the item at index and unlinks the list from the tensordict.

remove(value: Any) → None

Removes the first item equal to value and unlinks the list from the tensordict.

tolist() → list

Returns a plain list copy of the items, not linked to the tensordict.

Nested LinkedList items, as td[key] returns for a non-tensor entry with more than one batch dimension, become nested lists, as with tolist(). The other items are not copied.

Examples

>>> from tensordict import TensorDict
>>> td = TensorDict(batch_size=[2, 2])
>>> td["names"] = [["a", "b"], ["c", "d"]]
>>> td["names"].tolist()
[['a', 'b'], ['c', 'd']]