PyTorch torch.frexp function
Pytorch torch Reference Manual
torch.frexpis a function in PyTorch used to decompose floating-point numbers into mantissa and exponent.
Function Definition
torch.frexp(input)
Returns two tensors:
- mantissa - the mantissa part, in the range [-0.5, 1)
- exponent - the exponent part, an integer
Usage Example
Example
import torch
# Create a floating-point tensor
x = torch.tensor([8.0, 4.5, 2.0])
# Decompose into mantissa and exponent
mantissa, exponent = torch.frexp(x)
print("Mantissa:", mantissa)
print("Exponent:", exponent)
# Verify: x = mantissa * 2^exponent
print("Verification:", mantissa * 2 ** exponent)
# Create a floating-point tensor
x = torch.tensor([8.0, 4.5, 2.0])
# Decompose into mantissa and exponent
mantissa, exponent = torch.frexp(x)
print("Mantissa:", mantissa)
print("Exponent:", exponent)
# Verify: x = mantissa * 2^exponent
print("Verification:", mantissa * 2 ** exponent)
The output result is:
尾数: tensor([0.5000, 0.5625, 0.5000]) 指数: tensor([4, 3, 2]) 验证: tensor([8., 4.5, 2.])
Other Extensions