Function quantize

quantize(
    w: MlxArray,
    groupSize: number,
    bits: number,
    mode?: string,
    s?: bigint,
): QuantizedTensor
mx.quantize: w → (packed u32, scales, biases). mode defaults to the mlx-standard "affine" scheme; the quantizer seam may pass others.
Parameters
- w: MlxArray
- groupSize: number
- bits: number
- mode: string = "affine"
- s: bigint = gpuStream
Returns QuantizedTensor
- Defined in packages/mlx/src/ops.ts:745