For the complete documentation index, see llms.txt. Markdown versions of all pages are available by appending .md to any URL (e.g. /max/get-started.md).

Mojo function

fused_qk_rope_ragged

def fused_qk_rope_ragged[dtype: DType, freq_dtype: DType, collection_t: KVCollectionT, //, cache_t: KVCacheT, *, interleaved: Bool, target: StringSlice[StaticConstantOrigin], mrope_types: TypeList[mrope_types.values] = *?, mrope_section: Optional[Coord[mrope_types]] = None, PositionIdsLayoutType: TensorLayout = Layout[*?, *?]](q_proj: TileTensor[dtype, Storage=q_proj.Storage, address_space=q_proj.address_space, linear_idx_type=q_proj.linear_idx_type, element_size=q_proj.element_size], input_row_offsets: TileTensor[DType.uint32, Storage=input_row_offsets.Storage, address_space=input_row_offsets.address_space, linear_idx_type=input_row_offsets.linear_idx_type, element_size=input_row_offsets.element_size], kv_collection: collection_t, freqs_cis: TileTensor[freq_dtype, Storage=freqs_cis.Storage, address_space=freqs_cis.address_space, linear_idx_type=freqs_cis.linear_idx_type, element_size=freqs_cis.element_size], position_ids: OptionalReg[TileTensor[DType.uint32, PositionIdsLayoutType, ImmutAnyOrigin]], layer_idx: UInt32, output: TileTensor[dtype, Storage=output.Storage, address_space=output.address_space, linear_idx_type=output.linear_idx_type, element_size=output.element_size], context: DeviceContext)

Applies RoPE (Rotary Position Embedding) to query and key tensors.

This function can applies RoPE only to the last rope_dim elements of each head, leaving the first unroped_dim elements unchanged. This is required for DeepSeek models where only part of each head undergoes rotary transformation.