Is it possible to quantize less than 8-bit?

[jerryzh168 (Jerry Zhang)]

it’s not supported yet, responded here: Quantization aware training lower than 8-bits?