GPU tensorflow install issue -
i new user ubuntu, , spent more week trying install gpu tensorflow. tried follow tutorials website , blogs. whenever import tensorflow in python console. shows:
import tensorflow tensorflow/stream_executor/dso_loader.cc:128] opened cuda library libcublas.so.8.0 locally tensorflow/stream_executor/dso_loader.cc:128] opened cuda library libcudnn.so.5.1.5 locally tensorflow/stream_executor/dso_loader.cc:128] opened cuda library libcufft.so.8.0 locally tensorflow/stream_executor/dso_loader.cc:119] couldn't open cuda library libcuda.so.1. ld_library_path: :/usr/local/cuda/lib64:/usr/local/cuda/extras/cupti/lib64 tensorflow/stream_executor/cuda/cuda_diagnostics.cc:165] hostname: hanguo tensorflow/stream_executor/cuda/cuda_diagnostics.cc:189] libcuda reported version is: not found: unable find libcuda.so dso loaded program tensorflow/stream_executor/cuda/cuda_diagnostics.cc:193] kernel reported version is: permission denied: not open driver version path reading: /proc/driver/nvidia/version tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1091] ld_library_path: :/usr/local/cuda/lib64:/usr/local/cuda/extras/cupti/lib64 tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1092] failed find libcuda.so on system: failed precondition: not dlopen dso: libcuda.so.1; dlerror: libnvidia-fatbinaryloader.so.367.57: cannot open shared object file: no such file or directory tensorflow/stream_executor/dso_loader.cc:128] opened cuda library libcurand.so.8.0 locally
i tried search answer on google, no real solution came yet.can me out?
based on output $ ls /usr/local/cuda
, have installed successfully.
now, missing there path definition tensorflow can find is.
1) open .bashrc file: gedit ~/.bashrc
2) append following 2 lines end of file:
export ld_library_path="$ld_library_path:/usr/local/cuda/lib64:/usr/local/cuda/extras/cupti/lib64:/usr/local/cuda/include" export cuda_home=/usr/local/cuda
3) once save , close text file can return original terminal , type command reload .bashrc file:
source ~/.bashrc
although third part supposed take care of that, if want 100% sure work restart computer.
now, tensorflow, remember set cuda path /usr/local/cuda
, version 8.0.
Comments
Post a Comment