experimentalMax
fun experimentalMax(numLayers: Int, numHeads: Int, headDim: Int, maxSeqLen: Int): TurboQuantPreset(source)
Experimental maximum compression: TurboQuant-3 for both K and V.
Aggressive 3-bit compression. Use with caution — may degrade quality for some models. Best suited for long-context scenarios where memory is the primary constraint.