Variable failed to wrap HalfTensor

1

Is this a bug?

import torch

t = torch.HalfTensor([0])
t = torch.autograd.Variable(t)

This code causes these errors,

Traceback (most recent call last):
File “a.py”, line 4, in
t = torch.autograd.Variable(t)
RuntimeError: Variable data has to be a tensor, but got HalfTensor

2

Hi,
CPU half tensors do not actually exist.
Using cuda HalfTensors work as expected:

import torch

t = torch.cuda.HalfTensor([0])
t = torch.autograd.Variable(t)
3

Thank your for reply.

Are you intend to implement cpu HalfTensor?

I think It is desired torch.HalfTensor is deleted or sends warnings until it…
This is very unexpected behaviour.