mlx-bun library API
    Preparing search index...
    • Build trainable A/B leaves for the given ranks (one entry per module path). Modules absent from ranks are skipped. A is normal-init, B is zeros.

      Parameters

      • model: RuntimeModel
      • ranks: Map<string, number>
      • scale: number
      • seed: number
      • rsLora: boolean = false

      Returns TrainableLora