IMPORTANT: To view this page as Markdown, append `.md` to the URL (e.g. /get-started.md). For the complete documentation index, see llms.txt.
Skip to main content
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

PackedIndicesArgs

struct PackedIndicesArgs

The index build's operands, one kernel parameter.

Fields​

  • ​routing (In[.float32]): [tokens, 2 * top_k]: ids as float32, then weights.
  • ​order (Out[.uint32]): [slots] slot of each sorted position.
  • ​start (Out[.uint32]): [groups + 1] first sorted position of each group, then the total.
  • ​restore (Out[.uint32]): [slots] sorted position of each slot.
  • ​expert_ids (Out[.int32]): [groups] each group's id, g.
  • ​usage (Out[.uint32]): [2] the largest group's count, then the group count.
  • ​scales_offset (Out[.uint32]): [groups] scale-row tiles each group's start is shifted by.
  • ​weights (Out[.float32]): [slots] each slot's weight, dense.
  • ​tokens (Int64):
  • ​expert_offset (Int64):

Implemented traits​

AnyType, Copyable, Deinitable, DevicePassable, ImplicitlyCopyable, Movable

comptime members​

device_type​

comptime device_type = PackedIndicesArgs

Methods​

get_type_name​

static def get_type_name() -> String

Returns:

String

Was this page helpful?