51 Release the trained models in PyTorch format too, not just ONNX, for Quantization
Added support for quantized models in ModelRawBackend.
Added in8 quantized version of the ResNet-50-Arc model. This reduces the size of the model by a factor of 4 (86MB -> 21MB) and speeds up computation, at very little accuracy loss (further testing might be needed to ensure good accuracy, but preliminary testing yielded good results (no degradation)).
51 Release the trained models in PyTorch format too, not just ONNX, for Quantization