Skip to main content

gpu_maxpool2d

ferrotorch_gpu::kernels

Function gpu_maxpool2d 

Source 
pub fn gpu_maxpool2d(
    input: &CudaBuffer<f32>,
    batch: usize,
    channels: usize,
    h_in: usize,
    w_in: usize,
    kh: usize,
    kw: usize,
    sh: usize,
    sw: usize,
    ph: usize,
    pw: usize,
    device: &GpuDevice,
) -> GpuResult<(CudaBuffer<f32>, [usize; 4])>
Expand description

MaxPool2d forward on GPU. One thread per output element.