During annotation, we annotate each tensor as INPUT / WEIGHT / ACTIVATION kind for different quantization stratege, which is kinds of operation specific, ...
確定! 回上一頁