mlx-bun library API
    Preparing search index...
    • Quantize a model directory and write a loadable snapshot to outDir.

      Uniform affine N-bit by default. When opts.targetBpw is set, runs the OptiQ mixed-precision path (calibration → sensitivity → knapsack) to choose per-module bit-widths before writing.

      Parameters

      Returns Promise<QuantizeResult>