sparq
sparq copied to clipboard
modifing 8-bit int weight/activation
How can I access and modify the bit string of weight or activations of a CNN quantized model that can run on GPU?