Shift in conv2_8bit and conv_4bit

[jypark1257]

I understand how the conv2_2bits works,

but for conv2_4bits and conv_8bits, can you explain why there is "<<1" and "<<2" repectively in the function declaration?

Thank you

[alexmr09]

Hello,

The purpose of these shifts in the function declaration is to align the dimensions of the generated weight matrices after applying the Mixed Precision Quantization scheme for compression.

[jypark1257]

thank you