mstar.engine.resources.position.config

What a model declares about positions: its scheme, its spec, its step.

Kept free of the manager and its kernels so a submodule can declare a step without pulling FlashInfer in behind it.

Classes

PosBackend(*values)
PosScheme(*values)
PositionConfig(kv_cache, backend, scheme, ...)
PositionSpec(resource_key, nodes, config)
PositionStep([segments, pos_ids, advance])

class mstar.engine.resources.position.config.PosBackend(*values)

Bases: Enum

ROPE = 'rope'

class mstar.engine.resources.position.config.PosScheme(*values)

Bases: Enum

BLOCK = 'block'

SEQUENTIAL = 'sequential'

class mstar.engine.resources.position.config.PositionConfig(kv_cache: str, backend: mstar.engine.resources.position.config.PosBackend = <PosBackend.ROPE: 'rope'>, scheme: mstar.engine.resources.position.config.PosScheme = <PosScheme.SEQUENTIAL: 'sequential'>, block_step: int = 1, rotary_dim: int | None = None, interleave: bool = False, rope_scale: float = 1.0, rope_theta: float = 10000.0, rope_dtype: torch.dtype | None = None, low_freq_factor: float | None = None, high_freq_factor: float | None = None, old_context_len: int | None = None)

Bases: object

Parameters:

• kv_cache (str)

• backend (PosBackend)

• scheme (PosScheme)

• block_step (int)

• rotary_dim (int | None)

• interleave (bool)

• rope_scale (float)

• rope_theta (float)

• rope_dtype (dtype | None)

• low_freq_factor (float | None)

• high_freq_factor (float | None)

• old_context_len (int | None)

backend: PosBackend = 'rope'

block_step: int = 1

high_freq_factor: float | None = None

interleave: bool = False

kv_cache: str

property llama31_params: dict[str, float]

low_freq_factor: float | None = None

old_context_len: int | None = None

rope_dtype: dtype | None = None

rope_scale: float = 1.0

rope_theta: float = 10000.0

rotary_dim: int | None = None

scheme: PosScheme = 'sequential'

class mstar.engine.resources.position.config.PositionSpec(resource_key: str, nodes: set[str], config: mstar.engine.resources.position.config.PositionConfig)

Bases: NodeResourceSpec

Parameters:

• resource_key (str)

• nodes (set[str])

• config (PositionConfig)

config: PositionConfig

property resource_class: type[Resource]

What builds this spec. Imported inside the property, so declaring a resource stays free of the manager and its kernels.

The builder, not necessarily the class built: an attention spec names AttentionManager, whose build picks a backend subclass.

class mstar.engine.resources.position.config.PositionStep(segments: tuple[mstar.engine.resources.step.Segment, ...] | None = None, pos_ids: 'dict[str, torch.Tensor] | torch.Tensor | None' = None, advance: tuple[int, ...] | None = None)

Bases: ResourceStep

Parameters:

• segments (tuple[Segment, ...] | None)

• pos_ids (dict[str, Tensor] | Tensor | None)

• advance (tuple[int, ...] | None)

advance: tuple[int, ...] | None = None

pos_ids: dict[str, Tensor] | Tensor | None = None