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
paged_cache_rows
def paged_cache_rows[v_head_size: Int = Int(-1)](blocks: TileTensor[Engine=blocks.Engine, linear_idx_type=blocks.linear_idx_type], page_stride: TileTensor[.int64, Engine=page_stride.Engine, linear_idx_type=page_stride.linear_idx_type], cache_lengths: TileTensor[.uint32, Engine=cache_lengths.Engine, linear_idx_type=cache_lengths.linear_idx_type], lookup_table: TileTensor[.uint32, Engine=lookup_table.Engine, linear_idx_type=lookup_table.linear_idx_type], split_args: TileTensor[.uint32, Engine=split_args.Engine, linear_idx_type=split_args.linear_idx_type], first_row: Int, num_rows: Int, max_lengths_slot: Int, out result: PagedKVCacheCollection[blocks.dtype, _paged_cache_static_params(blocks.LayoutType.static_shape[Int(4)], blocks.LayoutType.static_shape[Int(5)], blocks.LayoutType.static_shape[Int(1)], v_head_size), blocks.LayoutType.static_shape[Int(3)], MutAnyOrigin, ImmutAnyOrigin, ImmutAnyOrigin, MutUntrackedOrigin])
Builds the paged cache over batch rows [first_row, first_row + num_rows), with max lengths from split_args[max_lengths_slot:+2].
Same type as the generic_get_paged_cache collection of the whole batch;
only the per-row tables move and the max lengths change. The caller
checks that the lookup table and split_args (a
kv_cache.mixed_step_split record) are packed. Use it to build a
collection from operands; rows_view views an existing collection and
keeps its scales.
Returns: