fp8_gemm_n128_row_scaled

metrale_model_layers::layers::ops

Function fp8_gemm_n128_row_scaled 

Source 
pub fn fp8_gemm_n128_row_scaled(
    gpu: &dyn GpuBackend,
    kernel: KernelHandle,
    input: DevicePtr,
    weight: &Fp8DenseWeight,
    output: DevicePtr,
    m: u32,
    n: u32,
    k: u32,
    stream: u64,
) -> Result<()>
Expand description

2026-09-25: Row-scaled FP8 GEMM, C[M, N] = A[M, K] @ (dequant(B_fp8[N, K]) * row_scale[N])^T; the scale multiplies each output column before the BF16 store. Takes the Fp8DenseWeight from crate::weight_map::DenseWeight::quantize_to_fp8. The DFlash drafter calls it when its quantization is Fp8Weights (dflash_head/forward_block_layer_paged.rs).

Kernel: fp8_gemm_t_row_scaled (per-model w4a16_gemm.cu).