xiguadong

> ONNX doesn't have a direct quantized tensor definition. Essentially, it uses QDQ to represent a quantized tensor. Thus we can limit the change to Q/DQ operators only as @daquexian...

> I think we would benefit from having a more structured discussion to consider the various aspects of representing quantization that are brought up in this thread. These aspects are...

I have met the same quesitions. So it seems like , when pads not specified and output_shape has specified，auto_pad=**NOTSET/SAME_UPPER** have same results.

> @xiguadong 编辑算子attribute的功能已经支持了：https://github.com/ZhangGe6/onnx-modifier#Edit_attribute_of_nodes 对于”插入自定义算子“，可否进一步描述一下需求？ 现在插入算子看起来都是在 onnx中已经定义好的，比如我有个自定义类型的Op `xigua` 想插入到模型中，应该是做不到的