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 #5693] Per-Model Concurrency #29308

New Issue
Closed
opened 2026-04-22 08:04:06 -05:00 by GiteaMirror · 3 comments
GiteaMirror commented 2026-04-22 08:04:06 -05:00
Owner
Copy Link

Originally created by @ProjectMoon on GitHub (Jul 15, 2024).
Original GitHub issue: https://github.com/ollama/ollama/issues/5693

I like the new concurrency features. I'm wondering if it would be possible to add a new Modelfile parameter to control parallel requests on a per-model basis. This would override OLLAMA_NUM_PARALLEL if set. The primary use-case for this idea is to allow small models like embedding models to serve many quick requests at once, while the larger models that take longer can serve a smaller number of requests at a time. This would allow larger models to be loaded more into the GPU, while the embedding models can work much faster.

Embedding creation from OpenWebUI is much much faster when hitting a bunch of documents (~45) with a high parallelism (20 in the test). However, this high parallelism forces the LLM model that will generate text to be mostly loaded into CPU, because it is also expecting to serve 20 parallel requests (when in reality it's going to serve one).

Originally created by @ProjectMoon on GitHub (Jul 15, 2024). Original GitHub issue: https://github.com/ollama/ollama/issues/5693 I like the new concurrency features. I'm wondering if it would be possible to add a new Modelfile parameter to control parallel requests on a per-model basis. This would override `OLLAMA_NUM_PARALLEL` if set. The primary use-case for this idea is to allow small models like embedding models to serve many quick requests at once, while the larger models that take longer can serve a smaller number of requests at a time. This would allow larger models to be loaded more into the GPU, while the embedding models can work much faster. Embedding creation from OpenWebUI is much much faster when hitting a bunch of documents (~45) with a high parallelism (20 in the test). However, this high parallelism forces the LLM model that will generate text to be mostly loaded into CPU, because it is also expecting to serve 20 parallel requests (when in reality it's going to serve one).
GiteaMirror added the feature request label 2026-04-22 08:04:07 -05:00
GiteaMirror commented 2026-04-22 08:04:09 -05:00
Author
Owner
Copy Link

@ProjectMoon commented on GitHub (Jul 15, 2024):

Based on quick skim of the code, it could perhaps be stuck in sched.go under processPending, and just set numParallel to the Modelfile's value, if it exists?

<!-- gh-comment-id:2227536626 --> @ProjectMoon commented on GitHub (Jul 15, 2024): Based on quick skim of the code, it could perhaps be stuck in sched.go under `processPending`, and just set `numParallel` to the Modelfile's value, if it exists?
GiteaMirror commented 2026-04-22 08:04:09 -05:00
Author
Owner
Copy Link

@ProjectMoon commented on GitHub (Jul 15, 2024):

https://github.com/ollama/ollama/pull/5657 also relevant.

<!-- gh-comment-id:2227538761 --> @ProjectMoon commented on GitHub (Jul 15, 2024): https://github.com/ollama/ollama/pull/5657 also relevant.
GiteaMirror commented 2026-04-22 08:04:10 -05:00
Author
Owner
Copy Link

@pdevine commented on GitHub (Sep 17, 2024):

This is definitely something that we're thinking about, but I'm going to close it as a dupe of #4894

cc @dhiltgen

<!-- gh-comment-id:2354340748 --> @pdevine commented on GitHub (Sep 17, 2024): This is definitely something that we're thinking about, but I'm going to close it as a dupe of #4894 cc @dhiltgen
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 feature request
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#29308
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?