Hi Grant, I have tried this several ways in context of our Watsonx challenge but in the end all ilab commands reported that it is still leveraging the CPU only and not my GPU. "ilab sysinfo" also claims "llama_cpp_python.supports_gpu_offload: False" no matter what I tried to get CUDA support into llama-cpp-python. The nvtop monitor only shows the activity of the screen recording and/or the browser. Please let me know how you checked that ilab is really leveraging the GPU because the method shown in the video does not work (at least for me 😞). Thanks for any response 👍
@matdavis137423 күн бұрын
I had the exact same experience, been struggling with this for a while and can't get it to work.