reference/generated/tensordict.TensorDictBase.tolist

Run in Google Colab

Colab

Download Notebook

Notebook

View on GitHub

GitHub

TensorDictBase.tolist

TensorDictBase.tolist(*, convert_nodes: bool = True, convert_tensors: bool | Literal['numpy'] = False, tolist_first: bool = False, as_linked_list: bool = False) → List[Any]

Returns the elements of the tensordict as nested lists, one level per batch dimension.

The innermost elements are the tensordict indexed at every batch index, converted with to_dict(): for batch size [2, 3], the result is a list of 2 lists of 3 dicts. A tensordict without batch dimensions returns one dict.

Keyword Arguments:

• convert_nodes (bool, optional) – if True, the elements are converted to dicts. If False, they are kept as tensordicts. Defaults to True.

• convert_tensors (bool or str, optional) – as in to_dict(): if True, tensors become Python lists, and if "numpy", NumPy arrays. Requires convert_nodes=True. Defaults to False.

• tolist_first (bool, optional) – as in to_dict(): if True, a nested tensordict with batch dimensions becomes a list of dicts. Defaults to False.

• as_linked_list (bool, optional) – if True, returns a LinkedList: assigning an element of the outer list (result[0] = {...}) writes it into the tensordict. Defaults to False.

Returns:

A list (nested once per batch dimension), or a dict (or self if convert_nodes=False) when the tensordict has no batch dimensions.

Raises:

TypeError – if convert_tensors is set and convert_nodes=False.

See also

to_dict().

Examples

>>> import torch
>>> from tensordict import TensorDict
>>> td = TensorDict(
...     a=torch.arange(4).view(2, 2),
...     b=TensorDict(c=torch.arange(8).view(2, 2, 2), batch_size=(2, 2, 2)),
...     batch_size=(2,),
... )
>>> td.tolist()
[{'a': tensor([0, 1]), 'b': {'c': tensor([[0, 1],
        [2, 3]])}}, {'a': tensor([2, 3]), 'b': {'c': tensor([[4, 5],
        [6, 7]])}}]
>>> td.tolist(convert_tensors=True)
[{'a': [0, 1], 'b': {'c': [[0, 1], [2, 3]]}}, {'a': [2, 3], 'b': {'c': [[4, 5], [6, 7]]}}]
>>> td.tolist(tolist_first=True, convert_tensors=True)[0]
{'a': [0, 1], 'b': [[{'c': 0}, {'c': 1}], [{'c': 2}, {'c': 3}]]}
>>> type(td.tolist(convert_nodes=False)[0]).__name__
'TensorDict'

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