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 #1389] Request: The ability to load multiple models into the same GPUs and running them concurrently. #62771

New Issue
Closed
opened 2026-05-03 10:16:51 -05:00 by GiteaMirror · 4 comments
GiteaMirror commented 2026-05-03 10:16:51 -05:00
Owner
Copy Link

Originally created by @phalexo on GitHub (Dec 5, 2023).
Original GitHub issue: https://github.com/ollama/ollama/issues/1389

Originally assigned to: @dhiltgen on GitHub.

Currently what ollama does is UNLOAD the previously loaded model, and loads the last model you try to use. Although the load is reasonably fast (if you intend to manually enter text and such) but if you want to use it with AutoGen or similar, loads and unloads put additional latency into the system, when token generation can already be pretty slow.

I am going try to separate GPUs into different groups and try to run different models within different groups, BUT it does not really solve the problem of resource utilization.

Thanks.

Originally created by @phalexo on GitHub (Dec 5, 2023). Original GitHub issue: https://github.com/ollama/ollama/issues/1389 Originally assigned to: @dhiltgen on GitHub. Currently what ollama does is UNLOAD the previously loaded model, and loads the last model you try to use. Although the load is reasonably fast (if you intend to manually enter text and such) but if you want to use it with AutoGen or similar, loads and unloads put additional latency into the system, when token generation can already be pretty slow. I am going try to separate GPUs into different groups and try to run different models within different groups, BUT it does not really solve the problem of resource utilization. Thanks.
GiteaMirror commented 2026-05-03 10:16:52 -05:00
Author
Owner
Copy Link

@phalexo commented on GitHub (Dec 5, 2023):

I was able to load different models into different sets of GPUs, but changing the ports manually, using OLLAMA_HOST,
was a pain. It should fail over to the next numbered port from the default 11434 to 11345, 11346,... if the port is already occupied.

<!-- gh-comment-id:1841354726 --> @phalexo commented on GitHub (Dec 5, 2023): I was able to load different models into different sets of GPUs, but changing the ports manually, using OLLAMA_HOST, was a pain. It should fail over to the next numbered port from the default 11434 to 11345, 11346,... if the port is already occupied.
GiteaMirror commented 2026-05-03 10:16:52 -05:00
Author
Owner
Copy Link

@shroominic commented on GitHub (Dec 24, 2023):

Would be sick to have it because it would enable 2 different models having a conversation without the delay of offloading the one and loading the other

<!-- gh-comment-id:1868599810 --> @shroominic commented on GitHub (Dec 24, 2023): Would be sick to have it because it would enable 2 different models having a conversation without the delay of offloading the one and loading the other
GiteaMirror commented 2026-05-03 10:16:54 -05:00
Author
Owner
Copy Link

@phalexo commented on GitHub (Dec 24, 2023):

You can do it now. I have been doing it. Just use multiple Ollama servers.

On Sun, Dec 24, 2023, 4:55 PM shroominic @.***> wrote:

Would be sick to have it because it would enable 2 different models having
a conversation without the delay of offloading the one and loading the other

—
Reply to this email directly, view it on GitHub
https://github.com/jmorganca/ollama/issues/1389#issuecomment-1868599810,
or unsubscribe
https://github.com/notifications/unsubscribe-auth/ABDD3ZOTLKQA4B4YV2KAN7LYLCQFTAVCNFSM6AAAAABAIBO4QKVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTQNRYGU4TSOBRGA
.
You are receiving this because you authored the thread.Message ID:
@.***>

<!-- gh-comment-id:1868600185 --> @phalexo commented on GitHub (Dec 24, 2023): You can do it now. I have been doing it. Just use multiple Ollama servers. On Sun, Dec 24, 2023, 4:55 PM shroominic ***@***.***> wrote: > Would be sick to have it because it would enable 2 different models having > a conversation without the delay of offloading the one and loading the other > > — > Reply to this email directly, view it on GitHub > <https://github.com/jmorganca/ollama/issues/1389#issuecomment-1868599810>, > or unsubscribe > <https://github.com/notifications/unsubscribe-auth/ABDD3ZOTLKQA4B4YV2KAN7LYLCQFTAVCNFSM6AAAAABAIBO4QKVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTQNRYGU4TSOBRGA> > . > You are receiving this because you authored the thread.Message ID: > ***@***.***> >
GiteaMirror commented 2026-05-03 10:16:54 -05:00
Author
Owner
Copy Link

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

Consolidating as a dup of #2109

<!-- gh-comment-id:1992109800 --> @dhiltgen commented on GitHub (Mar 12, 2024): Consolidating as a dup of #2109
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#62771
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?