mx.addmm(c, a, b): a@b + c fused (α=β=1) — nn.Linear's bias path (mx.addmm(bias, x, weight.T)).
mx.addmm(bias, x, weight.T)
mx.addmm(c, a, b): a@b + c fused (α=β=1) — nn.Linear's bias path (
mx.addmm(bias, x, weight.T)).