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 struct
RMSNormQuantizeDynamicScaledFloat8Dual
struct RMSNormQuantizeDynamicScaledFloat8Dual
Registers the mo.rms_norm_quantize_dynamic_scaled_float8.dual graph op: mo.reduce.rms_norm fused with a group-scaled (blockwise-FP8) mo.quantize_dynamic_scaled_float8, keeping the normalized row as a third output for the norm's other readers. Same outputs and scale layout as the unfused pair; group_size_or_per_token must be a proper divisor of the row.
Implemented traitsโ
Methodsโ
executeโ
static def execute[group_size_or_per_token: Int, multiply_before_cast: Bool, weight_offset: Float32](normed: T, output: T, scales: T, input: T, gamma: T, epsilon: Float32, scale_ub: Float32, ctx: DeviceContext)