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 struct

AMDStructuredConfig

struct AMDStructuredConfig[config: MHAConfig[config.dtype], group: Int, token_gen: Bool = False, mla_mode: Bool = False]

Implemented traits

AnyType, Copyable, ImplicitlyCopyable, ImplicitlyDeletable, Movable

comptime members

depth_padded

comptime depth_padded = False

double_buffer

comptime double_buffer = not token_gen

double_buffer_k_only

comptime double_buffer_k_only = token_gen and (config.block_n() <= 64)

full_kv

comptime full_kv = True

shared_kv

comptime shared_kv = token_gen and (config > 256)

Methods

heads_per_tile

static def heads_per_tile() -> Int

Returns:

Int

q_head_idx

static def q_head_idx() -> Int

Returns:

Int

q_tile_idx

static def q_tile_idx() -> Int

Returns:

Int

kv_head_idx

static def kv_head_idx() -> Int

Returns:

Int

get_mma_shape

static def get_mma_shape() -> IndexList[3]

Returns:

IndexList[3]

get_q_offset

static def get_q_offset[q_depth: Int]() -> UInt32

Returns:

UInt32

get_output_offset

static def get_output_offset[output_depth: Int]() -> UInt32

Returns:

UInt32