ByteTensor from boolean numpy array?

1

torch.ByteTensor(np.array([0,0,0,1,1,0],dtype=bool))

gives

184
87
241
165
91
127
[torch.ByteTensor of size 6]

I get similarly nonsensical results when converting to FloatTensor. Am I doing something obviously wrong, or is this a bug? Is anyone able to reproduce this?

Running this on pytorch 0.3.

2

My bad. This is a known issue. In case someone else has a similar problem, the workaround is to cast the numpy array as a supported dtype, eg uint8, prior to turning it into a ByteTensor.