skbio.tree.BPTree.numba_arrays#

BPTree.numba_arrays(gpu=None)[source]#

The tree’s arrays, for the Numba navigation functions.

Parameters:

gpumodule, optional: A Numba GPU module (numba_cuda_mlir.cuda or numba.cuda on NVIDIA, numba.hip on AMD) to place the arrays on a GPU: that of the tree’s data if it is on one, else the module’s current device. By default they stay on the host, for CPU code.

Returns:

tuple: An opaque bundle T of the tree’s topology and navigation index, passed to the functions of skbio.tree.bp.numba as their first argument, e.g. nav.parent(T, i). T.size is the number of parentheses.

See also

skbio.tree.bp.numba
lengths
edges

Notes

On the host, T holds read-only views of the tree’s own arrays: no copy is made. On a GPU, they are uploaded on the first call and cached on the tree, so later calls (and the batch operations with engine="numba") reuse them. This is so whether or not the tree was created from an array on the device.

The topology cannot change once the tree exists, so T stays valid as long as the tree. Node attributes can be replaced (e.g. with set_lengths), which is why they are not part of T: pass lengths or edges to a kernel as arguments of their own.

Examples

>>> from skbio.tree import BPTree
>>> tree = BPTree.read(["((a:1,b:2)c:3,(d:4,e:5)f:6)root;"])
>>> T = tree.numba_arrays()
>>> T.size
14