flashinfer.cake_fused_qk_rope_append.launch_plan

flashinfer.cake_fused_qk_rope_append.launch_plan(num_rows: int, num_requests: int, page_size: int, num_kv_heads: int, warps_per_row: int, clear_units_per_warp: int = 1) → dict[str, Any]

Grid and host-derived scalars (mirrors the Cake kernel’s launch_plan).

warps_per_row / clear_units_per_warp are the selected stage’s build knobs; the registry’s cake_jit.launch_plan() evaluates the same formula from a module record.