Will there be more examples of manually setting up QDQ (Quantize-Dequantize) in the future?

[lzcchl]

I used to use the pytorch_quantization library for QAT (Quantization-Aware Training). However, this library is no longer being maintained, and all of its content has been migrated to the modelopt library. I've noticed that there are not many examples available in the new library. I would like to ask the following questions:

How can I manually set up or configure QDQ (Quantize-Dequantize), and will future updates include the following examples:

1.How to initialize the quantization parameters of one node using those from another node;
just like (https://github.com/NVIDIA-AI-IOT/yolo_deepstream/blob/main/yolov7_qat/quantization/quantize.py#L219)

2.How to create a custom quantization class for a user-defined operation (custom OP);
just like (https://github.com/NVIDIA-AI-IOT/yolo_deepstream/blob/main/yolov7_qat/quantization/quantize.py#L46)

3.How to control certain layers so that they are excluded from quantization.
just like (https://github.com/NVIDIA-AI-IOT/yolo_deepstream/blob/main/yolov7_qat/quantization/quantize.py#L139)