Comment on Microsoft researchers developed a hyper-efficient AI model that can run on CPUsparentComments−imtringued1yAQLM, EfficientQAT and ParetoQ get reasonable benchmark scores at 2-bit quantization. At least 90% of the original unquantized scores.
Comments
AQLM, EfficientQAT and ParetoQ get reasonable benchmark scores at 2-bit quantization. At least 90% of the original unquantized scores.