Hi, I recently find that torch.Tensor
will automatically cast the input numpy.ndarray
to the float type. And I am wondering if there are anyway to bypass this.
For example:
import numpy as np, torch
a = np.arange(5).astype(np.int32)
b = torch.Tensor(a)
Then the tensor b
here will automatically get the type float
. I am aware that using b.int()
will cast b
into int, but I am wondering if there are any ways to avoid this by some mechanisms in torch.Tensor
.
Thank you!