reference/generated/tensordict.nn.inv_softplus

Run in Google Colab

Colab

Download Notebook

Notebook

View on GitHub

GitHub

inv_softplus

tensordict.nn.inv_softplus(bias: float | Tensor) → float | Tensor

Returns the input of the softplus function that gives bias: log(exp(bias) - 1).

exp(bias) - 1 is clamped to 1e-6 or more, so that the result is finite for bias <= 0.

Parameters:

bias (float or Tensor) – the output of the softplus function.

Returns:

A float for a float input, else a tensor.

Examples

>>> import torch
>>> from tensordict.nn import inv_softplus
>>> inv_softplus(torch.tensor([1.0, 2.0]))
tensor([0.5413, 1.8546])
>>> torch.nn.functional.softplus(inv_softplus(torch.tensor([1.0, 2.0])))
tensor([1., 2.])