1.芯片型号：X3
2.天工开物开发包OpenExplorer版本：XJ3_OE_2.4.2
3.问题定位：模型转换
4.问题具体描述：PTQ量化完，original.onnx推理都准确的，quantized.onnx最后几层的resize精度都非常底，请问是什么情况导致的，我转移到cpu上了,除此之外还有不少节点掉精度明显
Conv_5787 BPU id(0) HzSQuantizedConv 0.792077 0.351304 int8/int8
Conv_5856 BPU id(0) HzSQuantizedConv 0.978377 0.049981 int8/int8
Resize_5882 BPU id(0) HzQuantizedRoiResize 0.169139 0.579142 int8/int8
Resize_5895 BPU id(0) HzQuantizedRoiResize 0.077964 0.457910 int8/int8
Resize_5908 BPU id(0) HzQuantizedRoiResize 0.036449 0.306501 int8/int8
Resize_5921 BPU id(0) HzQuantizedRoiResize 0.015087 1.337093 int8/int8
UNIT_CONV_FOR_5904_0.01053_TO_FUSE_SCALE BPU id(0) HzSQuantizedConv int8/int8
UNIT_CONV_FOR_6866_0.01053_TO_FUSE_SCALE BPU id(0) HzSQuantizedConv int8/int8
UNIT_CONV_FOR_6879_0.01053_TO_FUSE_SCALE BPU id(0) HzSQuantizedConv int8/int8
UNIT_CONV_FOR_6892_0.01053_TO_FUSE_SCALE BPU id(0) HzSQuantizedConv int8/int8
Concat_5922 BPU id(0) Concat 0.957119 1.524482 int8/int8
Conv_5923 BPU id(0) HzSQuantizedConv 0.329454 1.337093 int8/int8
Conv_5926 BPU id(0) HzSQuantizedConv 0.982804 11.306360 int8/int32
Resize_5939 CPU – Resize 0.985186 – float/float
