[GH-ISSUE #8166] undefined symbol: ggml_backend_cuda_reg #5214

Closed
opened 2026-04-12 16:20:29 -05:00 by GiteaMirror · 2 comments
Owner

Originally created by @luisbrandao on GitHub (Dec 19, 2024).
Original GitHub issue: https://github.com/ollama/ollama/issues/8166

What is the issue?

I have 3 machines (2 rocky 9 and a rocky 8) All of then with nvidia GPUs

All of then are giving me the error:

$ ollama run llama3.1:8b
Error: llama runner process has terminated: error:/usr/local/lib/ollama/runners/cuda_v12_avx/ollama_llama_server: undefined symbol: ggml_backend_cuda_reg

On all versions after 0.5.1

i saw the same error on
https://github.com/ollama/ollama/issues/8149
https://github.com/ollama/ollama/issues/8038

But no clue how it was resolved.
What is happening? Did i miss some info?

I am using the install script.

OS

Linux

GPU

Nvidia

CPU

AMD

Ollama version

0.5.2~0.5.4

Originally created by @luisbrandao on GitHub (Dec 19, 2024). Original GitHub issue: https://github.com/ollama/ollama/issues/8166 ### What is the issue? I have 3 machines (2 rocky 9 and a rocky 8) All of then with nvidia GPUs All of then are giving me the error: ``` $ ollama run llama3.1:8b Error: llama runner process has terminated: error:/usr/local/lib/ollama/runners/cuda_v12_avx/ollama_llama_server: undefined symbol: ggml_backend_cuda_reg ``` On all versions after 0.5.1 i saw the same error on https://github.com/ollama/ollama/issues/8149 https://github.com/ollama/ollama/issues/8038 But no clue how it was resolved. What is happening? Did i miss some info? I am using the install script. ### OS Linux ### GPU Nvidia ### CPU AMD ### Ollama version 0.5.2~0.5.4
GiteaMirror added the bug label 2026-04-12 16:20:29 -05:00
Author
Owner

@luisbrandao commented on GitHub (Dec 19, 2024):

Ow, i get it now.

If anyone else have this problem, delete the entire thing first:
/usr/local/lib/ollama

then run the install script.

<!-- gh-comment-id:2552593732 --> @luisbrandao commented on GitHub (Dec 19, 2024): Ow, i get it now. If anyone else have this problem, delete the entire thing first: /usr/local/lib/ollama then run the install script.
Author
Owner

@snowpalm commented on GitHub (Jan 22, 2025):

Ow, i get it now.

If anyone else have this problem, delete the entire thing first: /usr/local/lib/ollama

then run the install script.

for me, it worked after deleting: /usr/lib/ollama (manual installation to upgrade from version 0.4x to 0.57, see the link to #8323 mentioned above in this thread). thanks anyway.

<!-- gh-comment-id:2606411636 --> @snowpalm commented on GitHub (Jan 22, 2025): > Ow, i get it now. > > If anyone else have this problem, delete the entire thing first: /usr/local/lib/ollama > > then run the install script. for me, it worked after deleting: /usr/lib/ollama (manual installation to upgrade from version 0.4x to 0.57, see the link to #8323 mentioned above in this thread). thanks anyway.
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#5214