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 struct
Struct_mla_prefill_graph_decode_paged_fp8_sparse
struct Struct_mla_prefill_graph_decode_paged_fp8_sparse
Registers the mo.mla.graph.prefill.decode.paged.fp8.sparse graph op with the graph compiler.
Implemented traits
Methods
execute
static def execute[dtype: DType, freq_dtype: DType, gamma_dtype: DType, fp8_dtype: DType, fp8_scale_dtype: DType, cache_dtype: DType, //, m_scale_granularity: Int, n_scale_granularity: Int, k_scale_granularity: Int, mask_str: StringSlice[ImmStaticOrigin], target: StringSlice[ImmStaticOrigin], indices_stride: Int, index_share: Bool = False](output: ManagedTensorSlice[IOSpec[_, _].Output, static_spec=output.static_spec], q: ManagedTensorSlice[IOSpec[_, _].Input, static_spec=q.static_spec], kv: ManagedTensorSlice[IOSpec[_, _].FusedInput, static_spec=kv.static_spec], input_row_offsets: ManagedTensorSlice[IOSpec[_, _].Input, static_spec=input_row_offsets.static_spec], freqs_cis: ManagedTensorSlice[IOSpec[_, _].Input, static_spec=freqs_cis.static_spec], kv_norm_gamma: ManagedTensorSlice[IOSpec[_, _].Input, static_spec=kv_norm_gamma.static_spec], buffer_row_offsets_1d: ManagedTensorSlice[IOSpec[_, _].Input, static_spec=buffer_row_offsets_1d.static_spec], cache_offsets_1d: ManagedTensorSlice[IOSpec[_, _].Input, static_spec=cache_offsets_1d.static_spec], buffer_length: Int32, w_k: ManagedTensorSlice[IOSpec[_, _].Input, static_spec=w_k.static_spec], w_uk: ManagedTensorSlice[IOSpec[_, _].Input, static_spec=w_uk.static_spec], w_uv: ManagedTensorSlice[IOSpec[_, _].Input, static_spec=w_uv.static_spec], kv_blocks: ManagedTensorSlice[IOSpec[_, _].MutableInput, static_spec=kv_blocks.static_spec], cache_lengths: ManagedTensorSlice[IOSpec[_, _].Input, static_spec=cache_lengths.static_spec], kv_lookup_table: ManagedTensorSlice[IOSpec[_, _].Input, static_spec=kv_lookup_table.static_spec], max_prompt_length: ManagedTensorSlice[IOSpec[_, _].Input, static_spec=max_prompt_length.static_spec], max_cache_length: ManagedTensorSlice[IOSpec[_, _].Input, static_spec=max_cache_length.static_spec], layer_idx: UInt32, scale: Float32, epsilon: Float32, w_k_scale: ManagedTensorSlice[IOSpec[_, _].Input, static_spec=w_k_scale.static_spec], w_uk_scale: ManagedTensorSlice[IOSpec[_, _].Input, static_spec=w_uk_scale.static_spec], w_uv_scale: ManagedTensorSlice[IOSpec[_, _].Input, static_spec=w_uv_scale.static_spec], scalar_args: ManagedTensorSlice[IOSpec[_, _].Input, static_spec=scalar_args.static_spec], sparse_indices: ManagedTensorSlice[IOSpec[_, _].Input, static_spec=sparse_indices.static_spec], topk_lengths: ManagedTensorSlice[IOSpec[_, _].Input, static_spec=topk_lengths.static_spec], attn_sink: ManagedTensorSlice[IOSpec[_, _].Input, static_spec=attn_sink.static_spec], num_partitions_scalar: ManagedTensorSlice[IOSpec[_, _].Input, static_spec=num_partitions_scalar.static_spec], context: DeviceContext)
Was this page helpful?
Thank you! We'll create more content like this.
Thank you for helping us improve!