aligned_array#

ray.rllib.utils.numpy.aligned_array(size: int, dtype, align: int = 64) → ndarray[source]#

Returns an array of a given size that is 64-byte aligned.

The returned array can be efficiently copied into GPU memory by TensorFlow.

Parameters:

size (int) – The size (total number of items) of the array. For example, array([[0.0, 1.0], [2.0, 3.0]]) would have size=4.
dtype – The numpy dtype of the array.
align (int) – The alignment to use.

Returns:

A np.ndarray with the given specifications.

Return type: