Open lgen opened 6 years ago
Ah, sorry to hear that. Are you explicitly passing the location of the cuda toolkit when you compile matconvnet? E.g. vl_compilenn(...., 'cudaRoot', 'path/to/your/cuda-8', ...)
?
Yes, i have tried, but it still doesn't work. Should i use a higher version of MATLAB ?
I'm not 100% certain to be honest, but it is worth a try if that is an option. Out of the box, I think 2015b is only guaranteed to be compatible with cuda 7.
When i compile matconvnet-1.0-beta25, i get the fellowing error:
警告: Version 7.0 of the CUDA toolkit could not be found. If installed, set MW_NVCC_PATH environment variable to location of nvcc compiler.
出错 mexcuda (line 157) [varargout{1:nargout}] = mex(mexArguments{:});
出错 vl_compilenn>mexcuda_compile (line 603) mexcuda(args{:}) ;
出错 vl_compilenn (line 489) mexcuda_compile(opts, srcs{i}, objfile, flags) ;
My system setting is ubuntu16.04,cuda8.0 ,matlab2015b. And i can compile matconvnet-1.0-beta24 and matconvnet-1.0-beta23 successfully. Could you give me some help about solving this problem ? Thank you.