Skip to main content

rms_norm_forward

teeny_kernels::nn::norm::rmsnorm

Function rms_norm_forward 

Source 
pub fn rms_norm_forward<T: Triton, D: Float, const BLOCK_N: i32>(
    x_ptr: T::Pointer<D>,
    y_ptr: T::Pointer<D>,
    weight_ptr: T::Pointer<D>,
    rrms_ptr: T::Pointer<D>,
    _M: i32,
    N: i32,
    eps: f32,
)
where
    T::I32Tensor: Tensor<i32, 1> + Comparison<i32, BoolTensor = T::BoolTensor>,
    T::Pointer<D>: AddOffsets<i32, 1, T::I32Tensor, Output = T::Tensor<T::Pointer<D>>>,
Expand description

RMSNorm forward pass.

Grid: [M] — one CTA per row.