l2_norm

Function l2_norm 

Source
pub fn l2_norm(
    gpu: &dyn GpuBackend,
    kernel: KernelHandle,
    data: DevicePtr,
    num_heads: u32,
    head_dim: u32,
    eps: f32,
    num_tokens: u32,
    stride: u32,
    stream: u64,
) -> Result<()>
Expand description

2026-09-25: In-place L2 normalization of each head, x / sqrt(sum(x^2) + eps) (kernel l2_norm_bf16). data is [num_tokens, stride] BF16 with num_heads * head_dim used per token; one block per (head, token).