github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-07 16:40:08 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #5397] V0.1.48 The model is loaded into the GPU Memory but runs on the CPU #3377

New Issue
Closed
opened 2026-04-12 14:00:30 -05:00 by GiteaMirror · 5 comments
GiteaMirror commented 2026-04-12 14:00:30 -05:00
Owner
Copy Link

Originally created by @wxtt522 on GitHub (Jul 1, 2024).
Original GitHub issue: https://github.com/ollama/ollama/issues/5397

Originally assigned to: @dhiltgen on GitHub.

What is the issue?

ollama run gemma2:27b
image
The same goes for loading other models. It was normal in the previous version. I did not change any environment variables.

OS

Windows

GPU

Nvidia

CPU

Intel

Ollama version

0.1.48

Originally created by @wxtt522 on GitHub (Jul 1, 2024). Original GitHub issue: https://github.com/ollama/ollama/issues/5397 Originally assigned to: @dhiltgen on GitHub. ### What is the issue? ollama run gemma2:27b ![image](https://github.com/ollama/ollama/assets/28422636/313df4c1-141f-4c61-aeeb-3525cc6fd975) The same goes for loading other models. It was normal in the previous version. I did not change any environment variables. ### OS Windows ### GPU Nvidia ### CPU Intel ### Ollama version 0.1.48
GiteaMirror added the bugneeds more infowindows labels 2026-04-12 14:00:30 -05:00
GiteaMirror commented 2026-04-12 14:00:32 -05:00
Author
Owner
Copy Link

@issaccv commented on GitHub (Jul 2, 2024):

I also encountered the same problem. According to my investigation, the problem is caused by OLLAMA_NUM_PARALLEL. When the value of this flag is set to a large value (such as 16 or 32), ollama will allocate more memory than VRAM, which causes some layers not to be offloaded to the GPU.

<!-- gh-comment-id:2201970011 --> @issaccv commented on GitHub (Jul 2, 2024): I also encountered the same problem. According to my investigation, the problem is caused by OLLAMA_NUM_PARALLEL. When the value of this flag is set to a large value (such as 16 or 32), ollama will allocate more memory than VRAM, which causes some layers not to be offloaded to the GPU.
GiteaMirror commented 2026-04-12 14:00:33 -05:00
Author
Owner
Copy Link

@dhiltgen commented on GitHub (Jul 2, 2024):

@wxtt522 can you share your server log so I can see what's going on? Can you also share what ollama ps says?

<!-- gh-comment-id:2204363209 --> @dhiltgen commented on GitHub (Jul 2, 2024): @wxtt522 can you share your server log so I can see what's going on? Can you also share what `ollama ps` says?
GiteaMirror commented 2026-04-12 14:00:34 -05:00
Author
Owner
Copy Link

@wxtt522 commented on GitHub (Jul 3, 2024):

@wxtt522 can you share your server log so I can see what's going on? Can you also share what ollama ps says?

server-1.log

Of course, this is my server log, I have adjusted the parameters mentioned above OLLAMA_NUM_PARALLEL, but it has no effect.
I noticed memory.available in the log，This is not normal. I found that the model started to be loaded into the memory when the gpu:p40 was loaded into the 4g gpu memory，But I didn't find any other way to adjust it

<!-- gh-comment-id:2205009363 --> @wxtt522 commented on GitHub (Jul 3, 2024): > @wxtt522 can you share your server log so I can see what's going on? Can you also share what `ollama ps` says? [server-1.log](https://github.com/user-attachments/files/16076630/server-1.log) Of course, this is my server log, I have adjusted the parameters mentioned above OLLAMA_NUM_PARALLEL, but it has no effect. I noticed memory.available in the log，This is not normal. I found that the model started to be loaded into the memory when the gpu:p40 was loaded into the 4g gpu memory，But I didn't find any other way to adjust it
GiteaMirror commented 2026-04-12 14:00:36 -05:00
Author
Owner
Copy Link

@wxtt522 commented on GitHub (Jul 3, 2024):

(base) PS C:\Users\admin> ollama ps
NAME ID SIZE PROCESSOR UNTIL
deepseek-coder-v2:16b-lite-instruct-q8_0 44250301ba51 18 GB 83%/17% CPU/GPU Forever

<!-- gh-comment-id:2205059495 --> @wxtt522 commented on GitHub (Jul 3, 2024): (base) PS C:\Users\admin> ollama ps NAME ID SIZE PROCESSOR UNTIL deepseek-coder-v2:16b-lite-instruct-q8_0 44250301ba51 18 GB 83%/17% CPU/GPU Forever
GiteaMirror commented 2026-04-12 14:00:37 -05:00
Author
Owner
Copy Link

@wxtt522 commented on GitHub (Jul 3, 2024):

Although I don’t know the reason, I solved the problem by resetting the graphics card driver configuration

<!-- gh-comment-id:2205283354 --> @wxtt522 commented on GitHub (Jul 3, 2024): Although I don’t know the reason, I solved the problem by resetting the graphics card driver configuration
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 needs more info windows
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#3377
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?