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
composite_layer_norm_rope_ragged_shape
def composite_layer_norm_rope_ragged_shape(input: T, gamma: T, beta: T, epsilon: Float32, input_row_offsets: T, start_pos: T, freqs_cis: T) -> IndexList[T.rank]
Computes the output shape for the mo.composite.layer_norm_rope_ragged graph op.
Args:
- input (
T): Activation tensor normalized by LayerNorm then partially rotated by ragged RoPE. - gamma (
T): Per-column scale weights applied after normalization. - beta (
T): Per-column shift weights applied after scaling. - epsilon (
Float32): Small constant added inside the normalization variance for numerical stability. - input_row_offsets (
T): Ragged batch boundaries. - start_pos (
T): Per-sequence cache length used for the RoPE position lookup. - freqs_cis (
T): RoPE frequency table; its width sets the rotated prefix.
Returns:
IndexList[T.rank]: The output shape, which matches the input shape.