bitorch_engine.layers.qlinear.nbit.mps.mpq_layer

Classes

MPQLinearMlx(*args, **kwargs)

Represents a MPS-compatible implementation of the mixed precision quantized (MPQ) linear layer, inheriting from MPQLinearBase.

MPQLinearMlxFunction(*args, **kwargs)

A custom autograd function for mixed-precision quantized (MPQ) linear operations on MPS acceleated by mlx.