How were the weights of these networks created? By post-training static quantization of the weights of the corresponding networks in torchvision.models? By quantization-aware training of the weights of the corresponding networks in torchvision.models? Thanks in advance!