How to generate C++ code for ResNet18 without optimizations?

[2212221352]

How can I generate C++ code for ResNet18 without the optimizations applied by ScaleHLS? Any suggestions?

[2212221352]

I appreciate someone helping me, and I will be very grateful.

[HahaLan97]

So, scaleflow-pytorch-pipeline can take serveral options, one of which is debug-point. If you take a look at line 159 in lib/Transforms/Passes.cpp, you can see when debug-point is equal to one, there would be no optimization applied.
If I understand the code correctly, you can simply use something like this to do get the unoptimized code:

scalehls-opt resnet18.mlir --scaleflow-pytorch-pipeline="top-func=forward debug-point=1" -o unoptimized.mlir

But I don't think you can get any C++ code from this because in this file everything is not properly lowered/converted yet. Hope this helps you.