For the complete documentation index, see llms.txt. Markdown versions of all pages are available by appending .md to any URL (e.g. /get-started.md).

Mojo function

rms_norm_fused_quantize_dynamic_scaled_fp8

def rms_norm_fused_quantize_dynamic_scaled_fp8[in_dtype: DType, out_dtype: DType, scale_dtype: DType, rank: Int, InputFn: def[width: Int, alignment: Int](Coord[*?]) -> SIMD[in_dtype, width] & RegisterPassable & ImplicitlyCopyable, OutputFn: def[width: SIMDLength, alignment: Int](Coord[*?], SIMD[out_dtype, width]) -> None & RegisterPassable & ImplicitlyCopyable, ScaleFn: def(Coord[*?], Scalar[scale_dtype]) -> None & RegisterPassable & ImplicitlyCopyable, AxisSizeT: CoordLike, /, target: StringSpan[ImmStaticOrigin], reduce_dim: Int = (rank - Int(1))](input_fn: InputFn, output_fn: OutputFn, scale_fn: ScaleFn, shape: Coord, axis_size: AxisSizeT, gamma: TileTensor[in_dtype, Engine=gamma.Engine, address_space=gamma.address_space, linear_idx_type=gamma.linear_idx_type], epsilon: Scalar[in_dtype], weight_offset: Scalar[in_dtype], scale_ub: Float32, context: Optional[DeviceContext] = None)