github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-07-15 20:54:54 -05:00
Code Issues 16.9k Packages Projects Releases 122 Wiki Activity

[GH-ISSUE #15956] Ollama 0.23 don't detect CUDA devices #87850

New Issue
Closed
opened 2026-05-10 06:27:00 -05:00 by GiteaMirror · 3 comments
GiteaMirror commented 2026-05-10 06:27:00 -05:00
Owner
Copy Link

Originally created by @trizt on GitHub (May 4, 2026).
Original GitHub issue: https://github.com/ollama/ollama/issues/15956

What is the issue?

OS: Artix Linux
Packages: ollama, ollama-cuda

When running ollama run llama3.2 "Count to 100" --verbose

0.23: total duration: 25.304998355s
llama_context: CPU output buffer size = 0.50 MiB

0.22: total duration: 2.639102603s
llama_context: CUDA_Host output buffer size = 0.50 MiB

I have two nVidia graphics cards, one is dedicated to a VM with vfio_pci and the other is the Linux machines graphics card which is used both for output and LLMs. Upgrade to the latest version everything runs on CPU, downgrading to 0.22 and things runs on the GPU (no reboot done).

Running nvtop shows no ollama among the GPU using processes listed when using version 0.23, while when running 0.22 there is the process listed and you also see memory and gpu usage.

discover/runner.go and ml/device.go had some changes, not sure if I can blame those or if it's something else that changed that causes the break for me at least.

Relevant log output

OS

Linux

GPU

Nvidia

CPU

AMD

Ollama version

0.23

Originally created by @trizt on GitHub (May 4, 2026). Original GitHub issue: https://github.com/ollama/ollama/issues/15956 ### What is the issue? OS: Artix Linux Packages: ollama, ollama-cuda When running **ollama run llama3.2 "Count to 100" --verbose** 0.23: total duration: 25.304998355s llama_context: CPU output buffer size = 0.50 MiB 0.22: total duration: 2.639102603s llama_context: CUDA_Host output buffer size = 0.50 MiB I have two nVidia graphics cards, one is dedicated to a VM with vfio_pci and the other is the Linux machines graphics card which is used both for output and LLMs. Upgrade to the latest version everything runs on CPU, downgrading to 0.22 and things runs on the GPU (no reboot done). Running nvtop shows no ollama among the GPU using processes listed when using version 0.23, while when running 0.22 there is the process listed and you also see memory and gpu usage. discover/runner.go and ml/device.go had some changes, not sure if I can blame those or if it's something else that changed that causes the break for me at least. ### Relevant log output ```shell ``` ### OS Linux ### GPU Nvidia ### CPU AMD ### Ollama version 0.23
GiteaMirror added the bug label 2026-05-10 06:27:00 -05:00
GiteaMirror commented 2026-05-10 06:27:02 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (May 4, 2026):

Server logs will aid in debugging.

<!-- gh-comment-id:4370540470 --> @rick-github commented on GitHub (May 4, 2026): [Server logs](https://docs.ollama.com/troubleshooting) will aid in debugging.
GiteaMirror commented 2026-05-10 06:27:03 -05:00
Author
Owner
Copy Link

@dquesadacr commented on GitHub (May 4, 2026):

I had the same issue. Somehow, the new ollama-cuda version does not provide libggml-cuda.so, found before at /usr/lib/ollama (on Artix/Arch). Fixed it by downgrading to v0.22.1-1, hope they fix it soon.

<!-- gh-comment-id:4370638872 --> @dquesadacr commented on GitHub (May 4, 2026): I had the same issue. Somehow, the new `ollama-cuda` version does not provide `libggml-cuda.so`, found before at `/usr/lib/ollama` (on Artix/Arch). Fixed it by downgrading to `v0.22.1-1`, hope they fix it soon.
GiteaMirror commented 2026-05-10 06:27:03 -05:00
Author
Owner
Copy Link

@trizt commented on GitHub (May 4, 2026):

Ollama 0.22.1 content in /usr/lib/ollama:
libggml-base.so libggml-base.so.0.0.0 libggml-cpu-haswell.so libggml-cpu-sandybridge.so libggml-cpu-sse42.so libggml-cuda.so
libggml-base.so.0 libggml-cpu-alderlake.so libggml-cpu-icelake.so libggml-cpu-skylakex.so libggml-cpu-x64.so

pacman -Ql ollama-cuda
ollama-cuda /usr/
ollama-cuda /usr/lib/
ollama-cuda /usr/lib/ollama/
ollama-cuda /usr/lib/ollama/libggml-cuda.so

Ollama 0.23.0 content in /usr/lib/ollama:
libggml-base.so libggml-base.so.0.0.0 libggml-cpu-haswell.so libggml-cpu-sandybridge.so libggml-cpu-sse42.so
libggml-base.so.0 libggml-cpu-alderlake.so libggml-cpu-icelake.so libggml-cpu-skylakex.so libggml-cpu-x64.so

pacman -Ql ollama-cuda
<no output at all>

so I can confirm that it's missing the libggml-cuda.so, it seems the ollama-cuda has become a empty package. I guess I have to go downstream with this issue, we can close it.

<!-- gh-comment-id:4371118449 --> @trizt commented on GitHub (May 4, 2026): Ollama 0.22.1 content in /usr/lib/ollama: libggml-base.so libggml-base.so.0.0.0 libggml-cpu-haswell.so libggml-cpu-sandybridge.so libggml-cpu-sse42.so libggml-cuda.so libggml-base.so.0 libggml-cpu-alderlake.so libggml-cpu-icelake.so libggml-cpu-skylakex.so libggml-cpu-x64.so pacman -Ql ollama-cuda ollama-cuda /usr/ ollama-cuda /usr/lib/ ollama-cuda /usr/lib/ollama/ ollama-cuda /usr/lib/ollama/libggml-cuda.so Ollama 0.23.0 content in /usr/lib/ollama: libggml-base.so libggml-base.so.0.0.0 libggml-cpu-haswell.so libggml-cpu-sandybridge.so libggml-cpu-sse42.so libggml-base.so.0 libggml-cpu-alderlake.so libggml-cpu-icelake.so libggml-cpu-skylakex.so libggml-cpu-x64.so pacman -Ql ollama-cuda `<no output at all>` so I can confirm that it's missing the libggml-cuda.so, it seems the ollama-cuda has become a empty package. I guess I have to go downstream with this issue, we can close it.
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
amd
api
app
bug
build
cli
cloud
compatibility
context-length
create
docker
documentation
embeddings
feature request
feedback wanted
good first issue
gpt-oss
gpu
harmony
help wanted
image
install
intel
js
launch
linux
macos
memory
mlx
model
needs more info
networking
nvidia
ollama.com
performance
pull-request
Mirrored from GitHub Pull Request
 python
question
registry
rendering
thinking
tools
top
vulkan
windows
wsl
No Label bug
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
GiteaMirror ninjasurge
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#87850
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?