Modular Arithmetic in PyTorch
Explanation
Modular arithmetic, also known as the remainder operation, is a function that returns the remainder when dividing $a$ by $b$. In PyTorch, there are two functions available:
torch.remainder(a,b)torch.fmod(a,b)
Both provide the remainder when $a$ is divided by $b$, but the outcomes are slightly different. If you’re curious about the specific formulas, refer to the official documents for remainder and fmod. Simply put, in remainder, the sign of the remainder is the same as that of the divisor $b$, whereas in mod, the sign of the remainder is the same as that of the dividend $a$.
>>> torch.remainder(torch.tensor([-3.1, 3.1, 1, -0.5, 2.1]), 2)
tensor([0.9000, 1.1000, 1.0000, 1.5000, 0.1000])
>>> torch.remainder(torch.tensor([-3.1, 3.1, 1, -0.5, 2.1]), -2)
tensor([-1.1000, -0.9000, -1.0000, -0.5000, -1.9000])
>>> torch.fmod(torch.tensor([-3.1, 3.1, 1, -0.5, 2.1]), 2)
tensor([-1.1000, 1.1000, 1.0000, -0.5000, 0.1000])
>>> torch.fmod(torch.tensor([-3.1, 3.1, 1, -0.5, 2.1]), -2)
tensor([-1.1000, 1.1000, 1.0000, -0.5000, 0.1000])
Environment
- OS: Windows11
- Version: Python v3.9.13, torch==1.13.1