dpctl.tensor.from_dlpack

dpctl.tensor.from_dlpack(obj)

Constructs dpctl.tensor.usm_ndarray instance from a Python object obj that implements __dlpack__ protocol. The output array is always a zero-copy view of the input.

Parameters:

obj – A Python object representing an array that supports __dlpack__ protocol.

Returns:

An array with a view into the tensor underlying the input obj.

Return type:

usm_ndarray

Raises:

• TypeError – if obj does not implement __dlpack__ method

• ValueError – if zero copy view can not be constructed because the input array resides on an unsupported device

See https://dmlc.github.io/dlpack/latest/ for more details.

Example:

import dpctl
import dpctl.tensor as dpt

class Container:
    "Helper class implementing `__dlpack__` protocol"
    def __init__(self, array):
        self._array = array

    def __dlpack__(self, stream=None):
        return self._array.__dlpack__(stream=stream)

    def __dlpack_device__(self):
        return self._array.__dlpack_device__()

C = Container(dpt.linspace(0, 100, num=20, dtype="int16"))
X = dpt.from_dlpack(C)