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 a nested list representation of the tensordict.

If the tensordict has no batch dimensions, this method returns a single list or dictionary. Otherwise, it returns a nested list where each inner list represents a batch dimension.

Parameters:

convert_nodes (bool) – if True, leaf nodes will be converted to dictionaries. Otherwise, they will be returned as tensordicts. Default: True.
convert_tensors (bool, "numpy") – if True, tensors will be converted to lists when creating the dictionary. If “numpy”, tensors will be converted to numpy arrays. Otherwise, they will remain as tensors. Default: False.
tolist_first (bool) – if True, the tensordict will be converted to a list first when it has batch dimensions. Default: False.
as_linked_list (bool) – if True, the list will be converted to a tensordict.utils.LinkedList which will automatically update the tensordict when the list is modified. Default: False.

Returns:

A nested list representation of the tensordict.

Examples

>>> import torch
>>> from tensordict import TensorDict
>>>
>>> td = TensorDict(
...     a=torch.arange(24).view(2, 3, 4),
...     b=TensorDict(c=torch.arange(12).reshape(2, 3, 2), batch_size=(2, 3, 2)),
...     batch_size=(2, 3)
... )
>>> print(td.tolist(tolist_first=True))
[[{'a': tensor([0, 1, 2, 3]), 'b': [{'c': tensor(0)}, {'c': tensor(1)}]}, {'a': tensor([4, 5, 6, 7]), 'b': [{'c': tensor(2)}, {'c': tensor(3)}]}, {'a': tensor([ 8,  9, 10, 11]), 'b': [{'c': tensor(4)}, {'c': tensor(5)}]}], [{'a': tensor([12, 13, 14, 15]), 'b': [{'c': tensor(6)}, {'c': tensor(7)}]}, {'a': tensor([16, 17, 18, 19]), 'b': [{'c': tensor(8)}, {'c': tensor(9)}]}, {'a': tensor([20, 21, 22, 23]), 'b': [{'c': tensor(10)}, {'c': tensor(11)}]}]]
>>> print(td.tolist(tolist_first=False))
[[{'a': tensor([0, 1, 2, 3]), 'b': {'c': tensor([0, 1])}}, {'a': tensor([4, 5, 6, 7]), 'b': {'c': tensor([2, 3])}}, {'a': tensor([ 8,  9, 10, 11]), 'b': {'c': tensor([4, 5])}}], [{'a': tensor([12, 13, 14, 15]), 'b': {'c': tensor([6, 7])}}, {'a': tensor([16, 17, 18, 19]), 'b': {'c': tensor([8, 9])}}, {'a': tensor([20, 21, 22, 23]), 'b': {'c': tensor([10, 11])}}]]
>>> print(td.tolist(convert_tensors=True))
[[{'a': [0, 1, 2, 3], 'b': {'c': [0, 1]}}, {'a': [4, 5, 6, 7], 'b': {'c': [2, 3]}}, {'a': [8, 9, 10, 11], 'b': {'c': [4, 5]}}], [{'a': [12, 13, 14, 15], 'b': {'c': [6, 7]}}, {'a': [16, 17, 18, 19], 'b': {'c': [8, 9]}}, {'a': [20, 21, 22, 23], 'b': {'c': [10, 11]}}]]
>>> print(td.tolist(convert_nodes=False)[0][0])
TensorDict(
    fields={
        a: Tensor(shape=torch.Size([4]), device=cpu, dtype=torch.int64, is_shared=False),
        b: TensorDict(
            fields={
                c: Tensor(shape=torch.Size([2]), device=cpu, dtype=torch.int64, is_shared=False)},
            batch_size=torch.Size([2]),
            device=None,
            is_shared=False)},
    batch_size=torch.Size([]),
    device=None,
    is_shared=False)

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

TensorDictBase.tolist#

Docs

Tutorials

Resources