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 #1521] Adjust Time for Ollama Serve to Stop Llama Runner Service #26589

New Issue
Closed
opened 2026-04-22 02:56:34 -05:00 by GiteaMirror · 3 comments
GiteaMirror commented 2026-04-22 02:56:34 -05:00
Owner
Copy Link

Originally created by @gzuuus on GitHub (Dec 14, 2023).
Original GitHub issue: https://github.com/ollama/ollama/issues/1521

Originally assigned to: @dhiltgen on GitHub.

Currently, the time it takes for Ollama Serve to stop the Llama Runner service is too short. It would be great to set the time to take longer to send the kill signal and stop the Llama Runner. Maybe its possible to add a configuration option to set the time it takes for Ollama Serve to stop the Llama Runner service.

Originally created by @gzuuus on GitHub (Dec 14, 2023). Original GitHub issue: https://github.com/ollama/ollama/issues/1521 Originally assigned to: @dhiltgen on GitHub. Currently, the time it takes for Ollama Serve to stop the Llama Runner service is too short. It would be great to set the time to take longer to send the kill signal and stop the Llama Runner. Maybe its possible to add a configuration option to set the time it takes for Ollama Serve to stop the Llama Runner service.
GiteaMirror commented 2026-04-22 02:56:35 -05:00
Author
Owner
Copy Link

@rgaidot commented on GitHub (Dec 14, 2023):

Yes, knowing that the server llama.cpp (example in their repo.) in permanent execution and much faster, no re-runs, etc..

I also understand why ollama wants to turn it off. In my opinion, because I haven't gone through all the code, if users want to switch models and ollama is using server from llama.cpp (*) with the old versions, you need to launch another server specifying the model for the /completion endpoint.

FYI, I'm using server from llama.cpp with their new endpoint (/v1/chat/completions).

(*) well, given the logs, there's a good chance

Best.

<!-- gh-comment-id:1856365753 --> @rgaidot commented on GitHub (Dec 14, 2023): Yes, knowing that the server llama.cpp (example in their repo.) in permanent execution and much faster, no re-runs, etc.. I also understand why ollama wants to turn it off. In my opinion, because I haven't gone through all the code, if users want to switch models and ollama is using server from llama.cpp (*) with the old versions, you need to launch another server specifying the model for the `/completion` endpoint. FYI, I'm using server from llama.cpp with their new endpoint (`/v1/chat/completions`). _(*) well, given the logs, there's a good chance_ Best.
GiteaMirror commented 2026-04-22 02:56:35 -05:00
Author
Owner
Copy Link

@transcriptionstream commented on GitHub (Dec 19, 2023):

I'm after similar, but opposite. The llama runner is holding gpu memory long after an api generate request is made. I'd love a way to adjust or stop the runners via api.

<!-- gh-comment-id:1863587757 --> @transcriptionstream commented on GitHub (Dec 19, 2023): I'm after similar, but opposite. The llama runner is holding gpu memory long after an api generate request is made. I'd love a way to adjust or stop the runners via api.
GiteaMirror commented 2026-04-22 02:56:35 -05:00
Author
Owner
Copy Link

@dhiltgen commented on GitHub (Mar 12, 2024):

In 0.1.25 we added keep_alive which allows you to control how long we keep the model loaded.

We've solved a number of memory ~leak problems, but are still working on some remaining ones which is why it never fully unloads from the GPU. That's tracked in issue #2767

<!-- gh-comment-id:1992138055 --> @dhiltgen commented on GitHub (Mar 12, 2024): In 0.1.25 we added `keep_alive` which allows you to control how long we keep the model loaded. We've solved a number of memory ~leak problems, but are still working on some remaining ones which is why it never fully unloads from the GPU. That's tracked in issue #2767
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
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#26589
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?