Converting tensorflow cast function with condition in pytorch | Kernel Crashes

vision
1

I’m looking for an equivalent of TensorFlow in Pytorch

binary_test = tf.cast(binary > 0.5, tf.float32)

I tried;

binary_test = binary.type(float32)
binary_test = torch.where(binary_test > 0.5, x, y)

where, x = Tensor([1]), y = Tensor([0])

On doing so the python kernel crashes.

2

The posted code should generally work:

binary = torch.randint(0, 2, (10,)).byte()
binary_test = binary.type(torch.float32)
binary_test = torch.where(binary_test > 0.5, torch.tensor([1.]), torch.tensor([0.]))

Note that I had to fix the float32 to torch.float32 and am not sure if it’s a copy-paste issue or if you might be running into this error.

Could you post the values of binary, which create this error please?