I’m new to TVM and I’m just wondering if it’ s possible to run the graph_runtime Debugger in C++? I’m working on a real time object detection application in C++, first results look promising but once I increase the input image sizes, inference time doubles compared to unoptimized version without TVM, I would like to dig into my network and check each layers execution time.
You might interested in How to Deploy TVM Modules for how to develop and deploy your program in C++. Specifically, for graph_runtime debugger you might need to get the registry "tvm.graph_runtime_debug.create" since it’s has been registered here.
Hope this helps.
Thanks @liangfu for your answer, getting "tvm.graph_runtime_debug.create" totally worked and calling the different methods in graph_runtime_debug.cc allowed me to find which layers are slowing back the inference when resizing the input image.