lukealonso

/

MiniMax-M2.7-NVFP4

like 37

Hey there,

Could you please explain how the quantization is done (is it modelopt) and how is this measured: "KLD reduced by ~10%." and is your calibration dataset on huggingface?

thanks!