github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-06 16:11:34 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #11474] 0.9.6 ollama can't find /usr/share/libdrm/amdgpu.ids instead if searches in /opt/amdgpu/share/libdrm/amdgpu.ids #69633

New Issue
Closed
opened 2026-05-04 18:43:23 -05:00 by GiteaMirror · 5 comments
GiteaMirror commented 2026-05-04 18:43:23 -05:00
Owner
Copy Link

Originally created by @ghost on GitHub (Jul 19, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/11474

What is the issue?

0.6.8 searches for /usr/share/libdrm/amdgpu.ids the correct one
While 0.9.6 uses /opt/amdgpu/share/libdrm/amdgpu.ids the non-existing one
Fedora 42.

Relevant log output

It basically says file not found `/opt/amdgpu/share/libdrm/amdgpu.ids`.
Both instances 0.6.8 and 0.9.6 are installed onto `/opt`

OS

Linux

GPU

AMD

CPU

AMD

Ollama version

0.9.6

Originally created by @ghost on GitHub (Jul 19, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/11474 ### What is the issue? 0.6.8 searches for /usr/share/libdrm/amdgpu.ids the correct one While 0.9.6 uses /opt/amdgpu/share/libdrm/amdgpu.ids the non-existing one Fedora 42. ### Relevant log output ```shell It basically says file not found `/opt/amdgpu/share/libdrm/amdgpu.ids`. Both instances 0.6.8 and 0.9.6 are installed onto `/opt` ``` ### OS Linux ### GPU AMD ### CPU AMD ### Ollama version 0.9.6
GiteaMirror added the amdbuglinux labels 2026-05-04 18:43:24 -05:00
GiteaMirror commented 2026-05-04 18:43:27 -05:00
Author
Owner
Copy Link

@ghost commented on GitHub (Jul 19, 2025):

Basically my RDNA2 laptop cannot load any models, because RoCM fails to load, I've setup the correct RDNA2 HSA_OVERRIDE that works with R7Pro 6850U iGPU.
But my RDNA3 desktop with older ollama version doesn't.

<!-- gh-comment-id:3092554114 --> @ghost commented on GitHub (Jul 19, 2025): Basically my RDNA2 laptop cannot load any models, because RoCM fails to load, I've setup the correct RDNA2 HSA_OVERRIDE that works with R7Pro 6850U iGPU. But my RDNA3 desktop with older ollama version doesn't.
GiteaMirror commented 2026-05-04 18:43:27 -05:00
Author
Owner
Copy Link

@dhiltgen commented on GitHub (Jul 31, 2025):

Can you share server logs with OLLAMA_DEBUG=1 set on both the old version that works and ideally the latest version (currently 0.10.1) so we can compare and see what might be going wrong?

<!-- gh-comment-id:3141594099 --> @dhiltgen commented on GitHub (Jul 31, 2025): Can you share server logs with OLLAMA_DEBUG=1 set on both the old version that works and ideally the latest version (currently 0.10.1) so we can compare and see what might be going wrong?
GiteaMirror commented 2026-05-04 18:43:28 -05:00
Author
Owner
Copy Link

@ghost commented on GitHub (Aug 6, 2025):

old.log
new.log
I think it works? gpu does get utilized.
I figure that on a laptop it puts a copy of the model into shared ram and then it's processed by cpu and gpu at the sametime, so I notice the GPU time utilization is off.

<!-- gh-comment-id:3158510791 --> @ghost commented on GitHub (Aug 6, 2025): [old.log](https://github.com/user-attachments/files/21614162/old.log) [new.log](https://github.com/user-attachments/files/21614822/new.log) I think it works? gpu does get utilized. I figure that on a laptop it puts a copy of the model into shared ram and then it's processed by cpu and gpu at the sametime, so I notice the GPU time utilization is off.
GiteaMirror commented 2026-05-04 18:43:28 -05:00
Author
Owner
Copy Link

@dhiltgen commented on GitHub (Aug 8, 2025):

It sounds like everything is working correctly. The new.log shows your GPU is discovered, and the model is loaded on it.

<!-- gh-comment-id:3168561381 --> @dhiltgen commented on GitHub (Aug 8, 2025): It sounds like everything is working correctly. The new.log shows your GPU is discovered, and the model is loaded on it.
GiteaMirror commented 2026-05-04 18:43:29 -05:00
Author
Owner
Copy Link

@ghost commented on GitHub (Aug 8, 2025):

Yeah I've looked it up, turns out the laptop wasn't loading all layers.
4GB dedicated vram wasn't enough to fit in a 4B Q4 model.

(...)
load_tensors: offloading 25 repeating layers to GPU
load_tensors: offloaded 25/37 layers to GPU
load_tensors:        ROCm0 model buffer size =  1517.35 MiB
load_tensors:   CPU_Mapped model buffer size =   976.34 MiB
(...)
llama_kv_cache_unified:      ROCm0 KV buffer size =   400.00 MiB
llama_kv_cache_unified:        CPU KV buffer size =   176.00 MiB

I just thought the model should have fit so the issue might have been bogus in a first place.

<!-- gh-comment-id:3169018662 --> @ghost commented on GitHub (Aug 8, 2025): Yeah I've looked it up, turns out the laptop wasn't loading all layers. 4GB dedicated vram wasn't enough to fit in a 4B Q4 model. ``` (...) load_tensors: offloading 25 repeating layers to GPU load_tensors: offloaded 25/37 layers to GPU load_tensors: ROCm0 model buffer size = 1517.35 MiB load_tensors: CPU_Mapped model buffer size = 976.34 MiB (...) llama_kv_cache_unified: ROCm0 KV buffer size = 400.00 MiB llama_kv_cache_unified: CPU KV buffer size = 176.00 MiB ``` I just thought the model should have fit so the issue might have been bogus in a first place.
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 amd bug linux
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#69633
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?