Package Details: ollama-cuda-git 0.1.30.gc2712b55-1

Git Clone URL: (read-only, click to copy)
Package Base: ollama-cuda-git
Description: Create, run and share large language models (LLMs) with CUDA
Upstream URL:
Licenses: MIT
Conflicts: ollama, ollama-cuda
Provides: ollama
Last Packager:
Votes: 1
Popularity: 0.49
First Submitted: 2024-02-22 23:22 (UTC)
Last Updated: 2024-04-01 12:42 (UTC)

Dependencies (4)

Required by (10)

Sources (4)

Latest Comments

nmanarch commented on 2024-04-17 07:49 (UTC) (edited on 2024-04-29 10:12 (UTC) by nmanarch)

I have found a little trick showed by others on ollama git issue. So for those want ollama cuda run without avx try :

Thanks to and to all.

Hello. I apologized. Since 1.29 the gpu support without avx cpu is blocked in ollama. Did someone can help to have this to work again? and the bypass propose by dbzoo which work but not apply at the main. Many thanks.have a nice days.

nmanarch commented on 2024-04-04 14:23 (UTC)

Yes it is fixed and build and run.thanks. commented on 2024-04-01 12:43 (UTC)

@nmanarch thanks for report. This problem must be fixed now

nmanarch commented on 2024-03-31 23:00 (UTC) (edited on 2024-03-31 23:59 (UTC) by nmanarch)

Hi ! the build failed..? someone have some tricks to solve ?

/var/tmp/pamac-build-nico/ollama-cuda-git/src/ollama/llm/llama.cpp/ note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 8, block_iq3_s, 1, &vec_dot_iq3_s_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI3_XS, block_iq3_s, 1, vec_dot_iq3_s_q8_1>
error: option 'cf-protection=return' cannot be specified on this target
error: option 'cf-protection=branch' cannot be specified on this target
194 warnings and 2 errors generated when compiling for gfx1010.
make[3]: *** [CMakeFiles/ggml.dir/build.make:132: CMakeFiles/ggml.dir/] Error 1
make[2]: *** [CMakeFiles/Makefile2:838: CMakeFiles/ggml.dir/all] Error 2
make[1]: *** [CMakeFiles/Makefile2:3575: ext_server/CMakeFiles/ext_server.dir/rule] Error 2
make: *** [Makefile:1440: ext_server] Error 2
llm/generate/generate_linux.go:3: running "bash": exit status 2
==> ERREUR : Une erreur s’est produite dans build().

So i have found on : So i remove my fcf-protection in my /etc/makepkg.conf .

But now i have

 g++ -fPIC -g -shared -o ../llama.cpp/build/linux/x86_64/rocm/lib/ -Wl,--whole-archive ../llama.cpp/build/linux/x86_64/rocm/ext_server/libext_server.a -Wl,--no-whole-archive ../llama.cpp/build/linux/x86_64/rocm/common/libcommon.a ../llama.cpp/build/linux/x86_64/rocm/libllama.a '-Wl,-rpath,$ORIGIN' -lpthread -ldl -lm -L/opt/rocm/lib -L/opt/amdgpu/lib/x86_64-linux-gnu/ '-Wl,-rpath,$ORIGIN/../../rocm/' -lhipblas -lrocblas -lamdhip64 -lrocsolver -lamd_comgr -lhsa-runtime64 -lrocsparse -ldrm -ldrm_amdgpu
/usr/sbin/ld : ../llama.cpp/build/linux/x86_64/rocm/ext_server/libext_server.a : membre %B dans l'archive n'est pas un objet
collect2: erreur: ld a retourné le statut de sortie 1
llm/generate/generate_linux.go:3: running "bash": exit status 1
==> ERREUR : Une erreur s’est produite dans build().