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
grouped_matmul_block_scaled_amd_kernel
def grouped_matmul_block_scaled_amd_kernel[out_dtype: DType, BLOCK_N: Int, c_layout: TensorLayout, a_layout: TensorLayout, b_layout: TensorLayout, a_scales_layout: TensorLayout, b_scales_layout: TensorLayout, row_offsets_layout: TensorLayout, expert_ids_layout: TensorLayout, CEngine: TensorEngine, AEngine: TensorEngine, BEngine: TensorEngine, AScalesEngine: TensorEngine, BScalesEngine: TensorEngine, RowOffsetsEngine: TensorEngine, ExpertIdsEngine: TensorEngine](c: TileTensor[out_dtype, c_layout, MutAnyOrigin, Engine=CEngine], a: TileTensor[.uint8, a_layout, ImmutAnyOrigin, Engine=AEngine], b: TileTensor[.uint8, b_layout, ImmutAnyOrigin, Engine=BEngine], a_scales: TileTensor[.float8_e8m0fnu, a_scales_layout, ImmutAnyOrigin, Engine=AScalesEngine], b_scales: TileTensor[.float8_e8m0fnu, b_scales_layout, ImmutAnyOrigin, Engine=BScalesEngine], row_offsets: TileTensor[.uint32, row_offsets_layout, ImmutAnyOrigin, Engine=RowOffsetsEngine], expert_ids: TileTensor[.int32, expert_ids_layout, ImmutAnyOrigin, Engine=ExpertIdsEngine], num_active_experts: Int32, M: Int32, N: Int32, K: Int32)