TensorDictBase.from_dict_instance#

TensorDictBase.from_dict_instance(input_dict, *, auto_batch_size: bool | None = None, batch_size=None, device=None, batch_dims=None, names=None)#

Builds a tensordict from a dictionary, with the types of the entries of this tensordict.

Works like from_dict(), except for the nested dicts: a nested dict whose key is a nested tensordict or tensorclass of this tensordict is converted to that type, recursively. The other nested dicts become TensorDict instances. Use it to turn the output of to_dict() back into tensorclasses. The values of this tensordict are not used, and the result is a TensorDict.

Parameters:

input_dict (dict) – the dictionary to convert. Nested dicts become nested entries.

Keyword Arguments:

auto_batch_size (bool, optional) – if True, sets the batch size to the largest leading shape that all the entries share. Defaults to None, which behaves like False unless batch_dims is given.
batch_size (Size or list of int, optional) – the batch size of the result: the leading dimensions that all its entries share. Defaults to None, which gives an empty batch size unless the batch size is computed.
device (device, str or int, optional) – the device of the result. Leaves on another device are moved to it. None leaves each leaf where it is, and the result has no device. Defaults to None.
batch_dims (int, optional) – with auto_batch_size, the maximum number of batch dimensions. Turns on auto_batch_size when that is None. Exclusive with batch_size. Defaults to None.
names (list of str, optional) – the names of the batch dimensions, one per dimension. Used only with batch_size: when the batch size is empty or computed, the names are dropped. Defaults to None.

Raises:

ValueError – if both batch_size and batch_dims are given.

Examples

>>> import torch
>>> from tensordict import TensorDict, tensorclass
>>> @tensorclass
... class MyClass:
...     x: torch.Tensor
>>> td = TensorDict(a=torch.zeros(()), b=MyClass(x=torch.zeros(())))
>>> data = {"a": torch.ones(()), "b": {"x": torch.ones(())}}
>>> type(td.from_dict_instance(data)["b"]).__name__
'MyClass'
>>> type(TensorDict.from_dict(data)["b"]).__name__
'TensorDict'

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

TensorDictBase.from_dict_instance#

Docs

Tutorials

Resources