warp-ctc icon indicating copy to clipboard operation
warp-ctc copied to clipboard

When installing, where to set cuda path?

Open xhran2010 opened this issue 6 years ago • 0 comments

My model can't run on GPU, I figured out it's an error during the installation. The document says:

Otherwise, set WARP_CTC_PATH to wherever you have libwarpctc.so installed. If you have a GPU, you should also make sure that CUDA_HOME is set to the home cuda directory (i.e. where include/cuda.h and lib/libcudart.so live). For example:

export CUDA_HOME="/usr/local/cuda"

But where should I set the CUDA_HOME??? Is it in the .bashrc? But I've already set! The model is still not running on GPU.

xhran2010 avatar Aug 04 '18 08:08 xhran2010