espnet.nets.pytorch_backend.nets_utils.to_torch_tensor

source

espnet.nets.pytorch_backend.nets_utils.to_torch_tensor(x)

Change to torch.Tensor or ComplexTensor from numpy.ndarray.

• Parameters:x – Inputs. It should be one of numpy.ndarray, Tensor, ComplexTensor, and dict.
• Returns: Type converted inputs.
• Return type: Tensor or ComplexTensor

Examples

>>> xs = np.ones(3, dtype=np.float32)
>>> xs = to_torch_tensor(xs)
tensor([1., 1., 1.])
>>> xs = torch.ones(3, 4, 5)
>>> assert to_torch_tensor(xs) is xs
>>> xs = {'real': xs, 'imag': xs}
>>> to_torch_tensor(xs)
ComplexTensor(
Real:
tensor([1., 1., 1.])
Imag;
tensor([1., 1., 1.])
)