bitorch_engine.layers.qlinear.nbit.cuda.mbwq_layer

Classes

MBWQLinearCuda(*args[, use_mbw, groups, ...])	Implements a Mixed-BitWidth Quantized (MBWQ) linear layer for CUDA devices.
MBWQLinearCudaFunction(*args, **kwargs)	Custom CUDA function for performing forward and backward passes in MBWQ Linear layers.