reference/generated/tensordict.TensorDictBase.from_dataclass

Run in Google Colab

Colab

Download Notebook

Notebook

View on GitHub

GitHub

TensorDictBase.from_dataclass

classmethod TensorDictBase.from_dataclass(dataclass, *, dest_cls: Type | None = None, auto_batch_size: bool = False, batch_dims: int | None = None, as_tensorclass: bool = False, device: device | None = None, batch_size: Size | None = None)

Converts a dataclass into a TensorDict instance.

Parameters:

dataclass – The dataclass instance to be converted.

Keyword Arguments:

• dest_cls (tensorclass, optional) – A tensorclass type to be used to map the data. If not provided, a new class is created. Without effect if obj is a type or as_tensorclass is False.

• auto_batch_size (bool, optional) – If True, automatically determines and applies batch size to the resulting TensorDict. Defaults to False.

• batch_dims (int, optional) – If auto_batch_size is True, defines how many dimensions the output tensordict should have. Defaults to None (full batch-size at each level).

• as_tensorclass (bool, optional) – If True, delegates the conversion to the free function from_dataclass() and returns a tensor-compatible class (tensorclass()) or instance instead of a TensorDict. Defaults to False.

• device (device, optional) – The device on which the TensorDict will be created. Defaults to None.

• batch_size (Size, optional) – The batch size of the TensorDict. Defaults to None.

Returns:

A TensorDict instance derived from the provided dataclass, unless as_tensorclass is True, in which case a tensor-compatible class or instance is returned.

Raises:

TypeError – If the provided input is not a dataclass instance.

Warning

This method is distinct from the free function from_dataclass and serves a different purpose. While the free function returns a tensor-compatible class or instance, this method returns a TensorDict instance.

Note

• This method creates a new TensorDict instance with keys corresponding to the fields of the input dataclass.

• Each key in the resulting TensorDict is initialized using the cls.from_any method.

• The auto_batch_size option allows for automatic batch size determination and application to the resulting TensorDict.

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