Closed Md7-jx closed 2 years ago
Which GPU are you using?
The problem is that gpu_cc.py fails to identify the model of GPU...
Which GPU are you using?
The problem is that gpu_cc.py fails to identify the model of GPU...
The GPU is GeForce RTX 3090.
I get something wrong with
libnames = ('libcuda.so', 'libcuda.dylib', 'cuda.dll') for libname in libnames: try: cuda = ctypes.CDLL(libname) except OSError: continue else: break else: return
There comes OSError,maybe I lost these 3 files.
My "libcuda.so" is located at /usr/lib/x86_64-linux-gnu/libcuda.so
. I think it was installed when I installed NVIDIA driver on the system.
Thanks for anwsering, I will try to update my NVIDIA driver .
Hi, Jkjung. Thanks for this project ,I succeeded it in ubuntu16.04 , but in Windows10 I failed. When I run "make" command under the folder "plugins/" ,I got computes: N o n e. "computes: N o n e NVCCFLAGS: -gencode arch=compute_N,code=[sm_N,compute_N] -gencode arch=compute_o,code=[sm_o,compute_o] -gencode arch=compute_n,code=[sm_n,compute_n] -gencode arch=compute_e,code=[sm_e,compute_e] nvcc -ccbin g++ -I"C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v11.1/include" -I"G:/test/TensorRT-8.2.4.2/include" -I"/usr/local/include" -I"plugin" -gencode arch=compute_N,code=[sm_N,compute_N] -gencode arch=compute_o,code=[sm_o,compute_o] -gencode arch=compute_n,code=[sm_n,compute_n] -gencode arch=compute_e,code=[sm_e,compute_e] -Xcompiler -fPIC -c -o yolo_layer.o yolo_layer.cu nvcc fatal : Unsupported gpu architecture 'compute_N' Makefile:43: recipe for target 'yolo_layer.o' failed mingw32-make: *** [yolo_layer.o] Error 1 " My gpu is RTX 3060. How to solve this problem?