Function HPGO::parallelism::gradient_accumulation::dp_cur_ga_p3_speedup

pub fn dp_cur_ga_p3_speedup(d: &Devices, m: &Model) -> f64