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
ihc_mix_gemv
def ihc_mix_gemv[k: Int, num_mix: Int, poison: Int = Int(0)](partials: TileTensor[.float32, Engine=partials.Engine, address_space=partials.address_space, linear_idx_type=partials.linear_idx_type], streams: TileTensor[Engine=streams.Engine, address_space=streams.address_space, linear_idx_type=streams.linear_idx_type], weight: TileTensor[.float32, Engine=weight.Engine, address_space=weight.address_space, linear_idx_type=weight.linear_idx_type], context: DeviceContext)
Launches the split-K iHC projection and sum of squares.
Parameters:
- k (
Int): Width of the flattened streams,hc_mult * hidden. - num_mix (
Int): Number ofweightrows. - poison (
Int): A planted fault (IHC_POISONS), 0 for none.
Args:
- partials (
TileTensor[.float32, Engine=partials.Engine, address_space=partials.address_space, linear_idx_type=partials.linear_idx_type]): Shape:[ihc_mix_gemv_splits[k](), num_rows, num_mix + 1]. - streams (
TileTensor[Engine=streams.Engine, address_space=streams.address_space, linear_idx_type=streams.linear_idx_type]): The flattened streams. Shape:[num_rows, k]. - weight (
TileTensor[.float32, Engine=weight.Engine, address_space=weight.address_space, linear_idx_type=weight.linear_idx_type]): The projection weight. Shape:[num_mix, k]. - context (
DeviceContext): The device context.
Raises:
If a shape disagrees with the parameters or the rows.