Open xfchangwei opened 2 years ago
Hi @xfchangwei
I guess this can be easily done with an environment variable:
export CUDA_VISIBLE_DEVICES=0
Hi @serizba I've tried this, but cppflow still doesn't seem to recognize my NVIDIA GPU. Is there cppflow example for using a specific GPU in multi GPU environments? For context, the C API of Tensorflow cannot find CUDA drivers, whereas the conda installed tensorflow package can use the GPU. Is this an issue with cppflow itself? Thanks!
How to specify GPU index? For example, I have multiple GPU devices