cupy.array — CuPy 13.4.1 documentation (original) (raw)

cupy.array(obj, dtype=None, copy=True, order='K', subok=False, ndmin=0, *, blocking=False)[source]#

Creates an array on the current device.

This function currently does not support the subok option.

Parameters:

• obj – cupy.ndarray object or any other object that can be passed to numpy.array().
• dtype – Data type specifier.
• copy (bool) – If False, this function returns obj if possible. Otherwise this function always returns a new array.
• order ({'C' , 'F' , 'A' , 'K'}) – Row-major (C-style) or column-major (Fortran-style) order. When order is 'A', it uses 'F' if a is column-major and uses 'C' otherwise. And when order is 'K', it keeps strides as closely as possible. If obj is numpy.ndarray, the function returns 'C'or 'F' order array.
• subok (bool) – If True, then sub-classes will be passed-through, otherwise the returned array will be forced to be a base-class array (default).
• ndmin (int) – Minimum number of dimensions. Ones are inserted to the head of the shape if needed.
• blocking (bool) – Default is False, meaning if a H2D copy is needed it would run asynchronously on the current stream, and users are responsible for ensuring the stream order. For example, writing to the source obj without proper ordering while copying would result in a race condition. If set to True, the copy is synchronous (with respect to the host).

Returns:

An array on the current device.

Return type:

cupy.ndarray

Note

This method currently does not support subok argument.

Note

If obj is an numpy.ndarray instance that contains big-endian data, this function automatically swaps its byte order to little-endian, which is the NVIDIA and AMD GPU architecture’s native use.