xnn.common.models.dispersion_fast.AUTO_POLICY

xnn.common.models.dispersion_fast.AUTO_POLICY = AutoPolicy(min_edges={}, min_edges_float64={}, default_min_edges=0, default_min_edges_float64=0)

minimum edges of the neighbor graph per GPU model. None: measured on A100, A30 and V100 in float32 and float64, the fast paths win at every size, from 1.2-2x on 24-atom clusters to 3-11x at 1536 atoms and 7-15x above 10k atoms

Type:

use_fast="auto"