Cmake couldnt find CUDA toolkit

[lament rivet]

tried every post related but not work

[tulip marsh]

I solve the problem with:
Uninstall Python3.11
Delete the Folder Python3.11
Delete VS Code Community.
Delete NVIDA Driver except grafic driver.
Reboot PC

Then Install:
-Python3.11
-VSCode Community

• Reboot PC
• Install NVIDA CUDA
  Use: https://github.com/KillianLucas/open-interpreter/blob/main/docs/GPU.md
  This install must find VS Studio Community (not VS CODE!)
  -Reboot PC
  Try with cmd:
  echo %CUDA_PATH%
  Must be like:
  C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.2
  Next command with cmd:
  set FORCE_CMAKE=1 && set CMAKE_ARGS=-DLLAMA_CUBLAS=on
  There is nothing to see. This is correct.
  Then try to install:

pip install llama-cpp-python --force-reinstall --upgrade --no-cache-dir

If is installed correct then go on:
After 12 hours trying to run interpreter i get after cmd:
python -c "from llama_cpp import GGML_USE_CUBLAS; print(GGML_USE_CUBLAS)"
a TRUE!!!!!!
I hope it helps a little bit.

[lament rivet]

wow ok

that seems like a full reinstall

what is the spec of the vs community

and reinstall vscode affect anything?

[lament rivet]

only a true without giving details about gpu. wierd. and gpu not working when running models.

[elfin fox]

this is defiantly not an option 😮