w4a16_gemv_silu_input

spark_model::layers::ops

Function w4a16_gemv_silu_input 

Source 
pub fn w4a16_gemv_silu_input(
    gpu: &dyn GpuBackend,
    kernel: KernelHandle,
    gate_out: DevicePtr,
    up_out: DevicePtr,
    weight: &QuantizedWeight,
    output: DevicePtr,
    n: u32,
    k: u32,
    stream: u64,
) -> Result<()>
Expand description

W4A16 GEMV with fused SiLU input: silu(gate)*up as activation, GEMV with down weights.

Reads gate_out[K] and up_out[K] BF16, computes silu(gate)*up per element inline, then multiplies by dequanted NVFP4 weights. Eliminates silu_mul kernel.

Grid: (ceil(N/4), 1, 1) Block: (256, 1, 1)