github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-07 00:22:43 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #10086] Long running ollama processes and high CPU usage #6611

New Issue
Closed
opened 2026-04-12 18:16:35 -05:00 by GiteaMirror · 4 comments
GiteaMirror commented 2026-04-12 18:16:35 -05:00
Owner
Copy Link

Originally created by @somera on GitHub (Apr 2, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/10086

What is the issue?

Since december 2024 I run Ollama in an Ubuntu VM on Proxmox. I started with Ollama v0.5.3 and now v0.6.3.

Sometimes I see that VRAM is used, but GPU only at ~10% and the CPU usage >70+%.

When this happens I see 1+ long running ollama process and RAM/VRAM usage.

Image

Image

or

Image

Image

or

Image

Image

In this case I restart the Ollama service.

Is this known problem?

What can I do to get more informations what happens here?

OS

Linux

GPU

Nvidia

CPU

AMD

Ollama version

0.6.3

Originally created by @somera on GitHub (Apr 2, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/10086 ### What is the issue? Since december 2024 I run Ollama in an Ubuntu VM on Proxmox. I started with Ollama v0.5.3 and now v0.6.3. Sometimes I see that VRAM is used, but GPU only at ~10% and the CPU usage >70+%. When this happens I see 1+ long running ollama process and RAM/VRAM usage. ![Image](https://github.com/user-attachments/assets/8c6aab76-3fdb-413e-80e1-d88deb2eca71) ![Image](https://github.com/user-attachments/assets/f8eed46e-bdda-487f-8206-6646929bb1e0) or ![Image](https://github.com/user-attachments/assets/eb4fb735-9792-40da-ac70-2e85f3aa4569) ![Image](https://github.com/user-attachments/assets/91734f59-29b4-4273-bc72-5b79f63dcb76) or ![Image](https://github.com/user-attachments/assets/0f35fd15-0690-4863-b829-144f6f431448) ![Image](https://github.com/user-attachments/assets/e1af6250-16bd-44c7-8295-ec643cec5b77) In this case I restart the Ollama service. Is this known problem? What can I do to get more informations what happens here? ### OS Linux ### GPU Nvidia ### CPU AMD ### Ollama version 0.6.3
GiteaMirror added the bug label 2026-04-12 18:16:35 -05:00
GiteaMirror commented 2026-04-12 18:16:37 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Apr 2, 2025):

If a model can't fit entirely in VRAM, part of the model is loaded in system RAM. ollama uses the GPU for inference for the part of the model in VRAM and the CPU for the part of the model in system RAM. The GPU is much faster than the CPU, so it completes its inference quickly. Then the CPU starts doing its inference, and it takes longer. This leads to low GPU utilization and high CPU utilization.

In some inferences, the model can become confused and continues to generate tokens without hitting an end-of-sequence token. This can lead to the situation where the model is in "Stopping..." state for a long time - the model is scheduled for unloading but ollama won't interrupt a model that's in the middle of an inference. This can be mitigated by setting num_predict in the API call or Modelfile.

<!-- gh-comment-id:2772000800 --> @rick-github commented on GitHub (Apr 2, 2025): If a model can't fit entirely in VRAM, part of the model is loaded in system RAM. ollama uses the GPU for inference for the part of the model in VRAM and the CPU for the part of the model in system RAM. The GPU is much faster than the CPU, so it completes its inference quickly. Then the CPU starts doing its inference, and it takes longer. This leads to low GPU utilization and high CPU utilization. In some inferences, the model can become confused and continues to generate tokens without hitting an end-of-sequence token. This can lead to the situation where the model is in "Stopping..." state for a long time - the model is scheduled for unloading but ollama won't interrupt a model that's in the middle of an inference. This can be mitigated by setting [`num_predict`](https://github.com/ollama/ollama/blob/main/docs/modelfile.md#valid-parameters-and-values:~:text=stop%20%22AI%20assistant%3A%22-,num_predict,-Maximum%20number%20of) in the API call or Modelfile.
GiteaMirror commented 2026-04-12 18:16:38 -05:00
Author
Owner
Copy Link

@somera commented on GitHub (Apr 2, 2025):

@rick-github thx for the response. I will test it.

<!-- gh-comment-id:2772006447 --> @somera commented on GitHub (Apr 2, 2025): @rick-github thx for the response. I will test it.
GiteaMirror commented 2026-04-12 18:16:38 -05:00
Author
Owner
Copy Link

@somera commented on GitHub (Apr 2, 2025):

In some inferences, the model can become confused and continues to generate tokens without hitting an end-of-sequence token. This can lead to the situation where the model is in "Stopping..." state for a long time - the model is scheduled for unloading but ollama won't interrupt a model that's in the middle of an inference.

Or you add an optional parameter to Ollama that allows you to say that if a model runs for longer than xx minutes/hours, it will be interrupted?

<!-- gh-comment-id:2772026918 --> @somera commented on GitHub (Apr 2, 2025): > In some inferences, the model can become confused and continues to generate tokens without hitting an end-of-sequence token. This can lead to the situation where the model is in "Stopping..." state for a long time - the model is scheduled for unloading but ollama won't interrupt a model that's in the middle of an inference. Or you add an optional parameter to Ollama that allows you to say that if a model runs for longer than xx minutes/hours, it will be interrupted?
GiteaMirror commented 2026-04-12 18:16:39 -05:00
Author
Owner
Copy Link

@somera commented on GitHub (Apr 28, 2025):

I'm closing this, cause this same with #10433

<!-- gh-comment-id:2836037937 --> @somera commented on GitHub (Apr 28, 2025): I'm closing this, cause this same with #10433
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#6611
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?