Convert Python Float to PyTorch Tensor - A Beginner Guide

Sometimes, we need to convert a python float data to pytorch tensor. In this tutorial, we will use some examples to show you how to convert.

Preliminary

Look at example code below:

import torch

f = 0.2

x = torch.cos(f)
print(x)

Run this code, we will see:

TypeError: cos(): argument ‘input’ (position 1) must be Tensor, not float

It means the python float data f is not a pytorch tensor. However, torch.cos() need a tensor input.

In order to fix this error, we need to convert python float data to pytorch tensor.

We can use torch.as_tensor() to convert.

For example:

import torch

f = 0.2

f = torch.as_tensor(f)
x = torch.cos(f)
print(x)

Run this code, we will see:

tensor(0.9801)

Morever, if you want to f support gradient, you can do as follows:

f = torch.as_tensor(f)
f = torch.nn.Parameter(f, requires_grad=True)
x = torch.cos(f)
print(x)

Here we use torch.nn.Parameter() to creat a parameter, then we can see:

tensor(0.9801, grad_fn=<CosBackward>)