Function loadKvCacheAsync

loadKvCacheAsync(
    path: string,
    model: { makeCache(): Cache[] },
    expect?: KvLoadExpect,
    codecs?: CacheCodecProvider,
): Promise<LoadedKvCache>
CPU-only asynchronous reads into unpublished allocations; no tensor copy crosses the event loop. Native array construction uses the shared codecs.
Parameters
- path: string
- model: { makeCache(): Cache[] }
- expect: KvLoadExpect = {}
- codecs: CacheCodecProvider = legacyCacheCodecs
Returns Promise<LoadedKvCache>
- Defined in packages/inference/src/state/persistence.ts:969