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
first_rejected_index
def first_rejected_index(rejected: TileTensor[.bool, Engine=rejected.Engine, address_space=rejected.address_space, linear_idx_type=rejected.linear_idx_type], dst: TileTensor[.int64, Engine=dst.Engine, address_space=dst.address_space, linear_idx_type=dst.linear_idx_type], ctx: DeviceContext)
Writes each row's first rejected draft position, steps if none.
Args:
- โrejected (
TileTensor[.bool, Engine=rejected.Engine, address_space=rejected.address_space, linear_idx_type=rejected.linear_idx_type]): Per-position rejection mask,[rows, steps]. - โdst (
TileTensor[.int64, Engine=dst.Engine, address_space=dst.address_space, linear_idx_type=dst.linear_idx_type]): The first rejected position per row,[rows, 1]. - โctx (
DeviceContext): The device context.
Raises:
Error: If the output does not carry exactly one slot per input row.