To compile the yolov4 models with batched NMS, I had to change this:
comment builder->allowGPUFallback(true); in SampleYolo.cpp. According docs the allowGPUFallback was removed from TensorRT 8.0.1. If I understand well, I do not have to care about it in TensorRT 8.0.1. Am I right?
comment all MYELIN_LIB and ENABLE_MYELIN, in Makefile.config to prevent error: /usr/bin/ld: cannot find -lmyelin because it seems to me that TensorRT 8.0.1 does not use it, but I found nothing about it. Do you have any idea?
If you do these 2 steps, then you can compile it and it works.
Good luck!
To compile the yolov4 models with batched NMS, I had to change this:
builder->allowGPUFallback(true);in SampleYolo.cpp. According docs theallowGPUFallbackwas removed from TensorRT 8.0.1. If I understand well, I do not have to care about it in TensorRT 8.0.1. Am I right?MYELIN_LIBandENABLE_MYELIN, inMakefile.configto prevent error:/usr/bin/ld: cannot find -lmyelinbecause it seems to me that TensorRT 8.0.1 does not use it, but I found nothing about it. Do you have any idea?If you do these 2 steps, then you can compile it and it works. Good luck!