Arch Linux

jronald

Member

Registered: 2022-05-09

Posts: 218

[SOLVED] How to install cuBLAS?

The command below doesn't install cuBLAS.

# pacman -S python pytorch-opt-cuda

BTW, when building whisper.cpp with WHISPER_CUBLAS on,
it says "cuBLAS not found".

Last edited by jronald (2023-10-09 15:13:29)

loqs

Member

Registered: 2014-03-06

Posts: 18,928

Re: [SOLVED] How to install cuBLAS?

How did you obtain whisper.cpp?  What is the full command you are using to build it?  What is the full output produced from the build command?
https://bbs.archlinux.org/viewtopic.php?id=57855

jronald

Member

Registered: 2022-05-09

Posts: 218

Re: [SOLVED] How to install cuBLAS?

How did you obtain whisper.cpp?  What is the full command you are using to build it?  What is the full output produced from the build command?
https://bbs.archlinux.org/viewtopic.php?id=57855

Using cmake GUI, default options with WHISPER_CUBLAS ON

loqs

Member

Registered: 2014-03-06

Posts: 18,928

Re: [SOLVED] How to install cuBLAS?

Works for me:

-- The C compiler identification is GNU 13.2.1
-- The CXX compiler identification is GNU 13.2.1
-- 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 "2.42.0") 
-- Performing Test CMAKE_HAVE_LIBC_PTHREAD
-- Performing Test CMAKE_HAVE_LIBC_PTHREAD - Success
-- Found Threads: TRUE  
-- Found CUDAToolkit: /opt/cuda/include (found version "12.2.91") 
-- cuBLAS found
-- The CUDA compiler identification is NVIDIA 12.2.91
-- Detecting CUDA compiler ABI info
-- Detecting CUDA compiler ABI info - done
-- Check for working CUDA compiler: /opt/cuda/bin/nvcc - skipped
-- Detecting CUDA compile features
-- Detecting CUDA compile features - done
-- CMAKE_SYSTEM_PROCESSOR: x86_64
-- x86 detected
-- GGML CUDA sources found, configuring CUDA architecture
-- Configuring done (1.8s)
-- Generating done (0.0s)

jronald

Member

Registered: 2022-05-09

Posts: 218

Re: [SOLVED] How to install cuBLAS?

As the logs show, "cuBLAS found", how to install cuBLAS?

Last edited by jronald (2023-10-09 12:37:04)

loqs

Member

Registered: 2014-03-06

Posts: 18,928

Re: [SOLVED] How to install cuBLAS?

Is it supplied by the cuda package which is a dependency of the package pytorch-opt-cuda you showed being installed.

jronald

Member

Registered: 2022-05-09

Posts: 218

Re: [SOLVED] How to install cuBLAS?

After rebooting, cuBLAS is found
Thanks