quantized models.
    *   Added support for converting and resizing models with dynamic
        (placeholder) dimensions. Previously, there was only limited support for
        dynamic batch size, and even that did not guarantee that the model could
        be properly resized at runtime.
        *   Enabled experimental support for a new quantization mode with 16-bit
            activations and 8-bit weights. See

GE,wEAAA,wEjJ08iDN,yDiJv8iDQ,iBAAA,eAGF,yCACE,WAAA,QAAA,wCAAA,mBACA,aAAA,QACA,MAAA,QAGF,0CAAA,0CAAA,wEAAA,wEAIE,WAAA,QAAA,wCAAA,mBACA,aAAA,QACA,MAAA,KAGF,4CAAA,4CAEE,iBAAA,eACA,aAAA,QACA,MAAA,QA9BN,6BACE,WAAA,QAAA,wCAAA,mBACA,MAAA,KAGE,sEAAA,sEjJs+iDN,uDiJn+iDQ,iBAAA,eAGF,uCACE,WAAA,QAAA,wCAAA,mBACA,aAAA,QACA,MAAA,QAGF,wCAAA,wCAAA,sEAAA,sEAIE,WAAA,QAAA,wCAAA,mBACA,aAAA,QACA,MAAA,KAGF,0CAAA,0CAEE,iBAAA,eACA,aAAA,QACA,MAAA,KA9BN,+BACE,WAAA,QAAA,wCAAA,mBACA,MAAA,KAGE,wEAAA,wEjJkgjDN,yDiJ//iDQ,iBAAA...