[−][src]Function HPGO::parallelism::gradient_accumulation::dp_ga_overlap_speedup
pub fn dp_ga_overlap_speedup(
d: &Devices,
m: &Model,
ga_size: u32,
inter_batch_overlap: bool
) -> f64
pub fn dp_ga_overlap_speedup(
d: &Devices,
m: &Model,
ga_size: u32,
inter_batch_overlap: bool
) -> f64