1.jit.save()的write failed问题 ：
模型保存路径不存在
2.FBGEMM doesn't support transpose packing yet!
set:
quantized.engine = 'qnnpack' and get_default_qconfig('qnnpack')
https://discuss.pytorch.org/t/is-there-a-way-to-quantize-conv-transpose2d-layer/96590/5
3.Could not export Python function call 'Scatter'.
TorchScript does not support tracing of Python functions
https://github.com/princeton-vl/CornerNet-Lite/issues/63
4.Could not run 'aten::add.Tensor' with arguments from the 'QuantizedCPU' backend
https://zhuanlan.zhihu.com/p/143664360
https://stackoverflow.com/questions/60519297/runtimeerror-could-not-run-atenxxxx-cant-load-pytorch-model-trained-with