AttentionBackend

class torchao.prototype.attention.AttentionBackend(value, names=None, *, module=None, qualname=None, type=None, start=1, boundary=None)

Backend kernel for computing attention.