Class ExpertIOSlabStore

Index

Constructors

Properties

files slotBytes slots

Methods

cancel close compressedMemory discard hintSegments hintTelemetry lease physicalFootprint pointer releaseAfterGpuSync releaseCpu releaseGpuFenced submit submitSegments view wait

Constructors

constructor

new ExpertIOSlabStore(
path: string | readonly string[],
options: ExpertIOOptions,
): ExpertIOSlabStore
Parameters
- path: string | readonly string[]
- options: ExpertIOOptions
Returns ExpertIOSlabStore
- Defined in packages/inference/src/artifacts/experts/io.ts:85

Properties

`Readonly`files

files: readonly string[]

`Readonly`slotBytes

slotBytes: number

`Readonly`slots

slots: number

Methods

cancel

cancel(slot: number, generation: bigint): void
Parameters
- slot: number
- generation: bigint
Returns void
- Defined in packages/inference/src/artifacts/experts/io.ts:277

close

close(): void
Returns void
- Defined in packages/inference/src/artifacts/experts/io.ts:344

compressedMemory

compressedMemory(): number
Returns number
- Defined in packages/inference/src/artifacts/experts/io.ts:325

discard

discard(slot: number, generation: bigint): void
Parameters
- slot: number
- generation: bigint
Returns void
- Defined in packages/inference/src/artifacts/experts/io.ts:314

hintSegments

hintSegments(segments: readonly ExpertIOHintSegment[]): boolean
Queue advisory readahead on the dedicated hint worker. No slab is used.
Parameters
- segments: readonly ExpertIOHintSegment[]
Returns boolean
- Defined in packages/inference/src/artifacts/experts/io.ts:223

hintTelemetry

hintTelemetry(): ExpertIOHintSnapshot
Returns ExpertIOHintSnapshot
- Defined in packages/inference/src/artifacts/experts/io.ts:255

lease

lease(slot: number, generation: bigint, consumer?: "gpu" | "cpu"): void
Parameters
- slot: number
- generation: bigint
- consumer: "gpu" | "cpu" = "cpu"
Returns void
- Defined in packages/inference/src/artifacts/experts/io.ts:283

physicalFootprint

physicalFootprint(): number
Returns number
- Defined in packages/inference/src/artifacts/experts/io.ts:320

pointer

pointer(slot: number, generation: bigint): number
Parameters
- slot: number
- generation: bigint
Returns number
- Defined in packages/inference/src/artifacts/experts/io.ts:330

releaseAfterGpuSync

releaseAfterGpuSync(
    slot: number,
    generation: bigint,
    dependentOutputs: readonly MlxArray[],
    stream?: bigint,
): Promise<void>
Parameters
- slot: number
- generation: bigint
- dependentOutputs: readonly MlxArray[]
- Optionalstream: bigint
Returns Promise<void>
- Defined in packages/inference/src/artifacts/experts/io.ts:301

releaseCpu

releaseCpu(slot: number, generation: bigint): void
Parameters
- slot: number
- generation: bigint
Returns void
- Defined in packages/inference/src/artifacts/experts/io.ts:289

releaseGpuFenced

releaseGpuFenced(slot: number, generation: bigint): void
Parameters
- slot: number
- generation: bigint
Returns void
- Defined in packages/inference/src/artifacts/experts/io.ts:295

submit

submit(slot: number, generation: bigint, offset: number, length: number): void
Parameters
- slot: number
- generation: bigint
- offset: number
- length: number
Returns void
- Defined in packages/inference/src/artifacts/experts/io.ts:150

submitSegments

submitSegments(
    slot: number,
    generation: bigint,
    segments: readonly ExpertIOSegment[],
): void
Parameters
- slot: number
- generation: bigint
- segments: readonly ExpertIOSegment[]
Returns void
- Defined in packages/inference/src/artifacts/experts/io.ts:159

view

view(slot: number, generation: bigint, length?: number): Uint8Array
Parameters
- slot: number
- generation: bigint
- Optionallength: number
Returns Uint8Array
- Defined in packages/inference/src/artifacts/experts/io.ts:337

wait

wait(slot: number, generation: bigint): Promise<void>
Parameters
- slot: number
- generation: bigint
Returns Promise<void>
- Defined in packages/inference/src/artifacts/experts/io.ts:215