Function bindEmbeddingsInput

bindEmbeddingsInput(
    prefill: (
        ids: MlxArray,
        caches: Cache[],
        start: number,
        phase: "tail" | "decode",
        width: number,
    ) => MlxArray,
): MlxPromptInput
Preserve the ordinary generator's projection geometry for a prepared embedding sequence: process the whole prompt, then project only its tip.
Parameters
- prefill: (
      ids: MlxArray,
      caches: Cache[],
      start: number,
      phase: "tail" | "decode",
      width: number,
  ) => MlxArray
Returns MlxPromptInput
- Defined in packages/inference/src/models/media-input.ts:7