Help to install llama-cpp-python binding on Ubuntu

[celsowm]

I am trying to install llama cpp on Ubuntu 23.10 using:

CMAKE_ARGS="-DLLAMA_CUBLAS=on" pip install llama-cpp-python

But I got this error:

[46 lines of output]
      *** scikit-build-core 0.7.1 using CMake 3.27.4 (wheel)
      *** Configuring CMake...
      loading initial cache file /tmp/tmpnqf71w1p/build/CMakeInit.txt
      -- The C compiler identification is GNU 13.2.0
      -- The CXX compiler identification is GNU 13.2.0
      -- Detecting C compiler ABI info
      -- Detecting C compiler ABI info - done
      -- Check for working C compiler: /usr/bin/cc - skipped
      -- Detecting C compile features
      -- Detecting C compile features - done
      -- Detecting CXX compiler ABI info
      -- Detecting CXX compiler ABI info - done
      -- Check for working CXX compiler: /usr/bin/c++ - skipped
      -- Detecting CXX compile features
      -- Detecting CXX compile features - done
      -- Found Git: /usr/bin/git (found version "2.40.1")
      -- Performing Test CMAKE_HAVE_LIBC_PTHREAD
      -- Performing Test CMAKE_HAVE_LIBC_PTHREAD - Success
      -- Found Threads: TRUE
      -- Unable to find cuda_runtime.h in "/usr/local/include" for CUDAToolkit_INCLUDE_DIR.
      -- Unable to find cublas_v2.h in either "" or "/math_libs/include"
      -- Unable to find cudart library.
      -- Could NOT find CUDAToolkit (missing: CUDAToolkit_INCLUDE_DIR CUDA_CUDART) (found version "11.8.89")
      CMake Warning at vendor/llama.cpp/CMakeLists.txt:360 (message):
        cuBLAS not found

Its crazy because it says "found version "11.8.89" about CUDA !

Any hints?