reference/generated/tensordict.TensorDictBase.clamp

Run in Google Colab

Colab

Download Notebook

Notebook

View on GitHub

GitHub

TensorDictBase.clamp

TensorDictBase.clamp(min: TensorDictBase | Tensor | float | None = None, max: TensorDictBase | Tensor | float | None = None, *, out=None) → Self

Returns a new tensordict with every leaf clamped into the range [min, max].

\[y_i = \min(\max(x_i, \text{min}_i), \text{max}_i)\]

Where min is greater than max, the result is max, as with torch.clamp(). At least one of min and max must be given; with only one bound, clamp calls clamp_min() or clamp_max().

Parameters:

• min (TensorDictBase, Tensor, Number or None, optional) – the lower bound. A tensordict is paired with the tensordict key by key. min and max must both be tensordicts, or neither. Defaults to None, which sets no lower bound.

• max (TensorDictBase, Tensor, Number or None, optional) – the upper bound. min and max must both be tensordicts, or neither. Defaults to None, which sets no upper bound.

Keyword Arguments:

out (TensorDictBase, optional) – the tensordict to write the result into, and the one returned. The result is copied into its leaves, so out must have the keys, shapes and dtypes of the result. Defaults to None.

Raises:

• ValueError – if one of min and max is a tensordict and the other is not, or if out is given and both min and max are None.

• KeyError – if out does not have the keys of the result.

• RuntimeError – if a leaf of out does not have the dtype of the result, as with torch.clamp().

Examples

>>> import torch
>>> from tensordict import TensorDict
>>> td = TensorDict(x=torch.tensor([-2.0, 0.5, 3.0]), batch_size=[3])
>>> td.clamp(0.0, 1.0)["x"]
tensor([0.0000, 0.5000, 1.0000])
>>> td.clamp(max=0.0)["x"]
tensor([-2.,  0.,  0.])

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