PyTorch .float_power()

In PyTorch, .float_power() raises each element of a tensor to the power of the given exponent with the result always returned in floating point.

Syntax

torch.float_power(input, exponent, *, out=None)

Parameters:

• input (Tensor): The base tensor.
• exponent (Tensor or Scalar): The exponent to which each element of the input is raised. Must be broadcastable to input.
• out (Tensor, optional): The output tensor to store the result.

Return value:

• Returns a tensor containing the input elements raised to the specified power, with the result always in the floating point data type.

Example 1: Tensor Base with Scalar Exponent

In this example, each element of the tensor is raised to the same scalar exponent 2:

import torch

base = torch.tensor([1, 2, 3, 4])
exp = 2
result = torch.float_power(base, exp)
print(result)

Copy to clipboard
Copy to clipboard

The output of this code is:

tensor([ 1.,  4.,  9., 16.])

Copy to clipboard
Copy to clipboard

Example 2: Tensor Base with Tensor Exponents

In this example, each element of the base tensor is raised to the corresponding element in the exponent tensor, allowing for fractional powers:

import torch

base = torch.tensor([1, 2, 3, 4])
exp = torch.tensor([0.5, 1, 1.5, 2])
result = torch.float_power(base, exp)
print(result)

Copy to clipboard
Copy to clipboard

The output of this code is:

tensor([ 1.0000,  2.0000,  5.1962, 16.0000])

Copy to clipboard
Copy to clipboard