bitorch_engine.layers.qlinear.nbit.cuda.utils

Functions

make_group_map	Creates a mapping of quantization groups for handling irregular group sizes in quantized models.
pack_fp_weight	Packs the fp16 weight into a quantized weight format using the attributes defined in the QweightParameter.
unpack_qweight	Reconstructs the fp16 weight tensor from the input quantized weight parameter.