pub fn moe_silu_mul(
gpu: &dyn GpuBackend,
kernel: KernelHandle,
gate: DevicePtr,
up: DevicePtr,
output: DevicePtr,
total_elements: u32,
stream: u64,
) -> Result<()>Expand description
Element-wise SiLU activation + multiply: output[i] = silu(gate[i]) * up[i].
Grid: (ceil(total_elements/256), 1, 1) Block: (256, 1, 1)