github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-07-15 12:48:05 -05:00
Code Issues 16.9k Packages Projects Releases 122 Wiki Activity

[GH-ISSUE #8947] How to enable data-parallel inference with multi-GPU? #5804

New Issue
Closed
opened 2026-04-12 17:08:48 -05:00 by GiteaMirror · 2 comments
GiteaMirror commented 2026-04-12 17:08:48 -05:00
Owner
Copy Link

Originally created by @lixiangge on GitHub (Feb 8, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/8947

I’m currently using Ollama with docker to deploy Qwen2.5 on a single GPU, and it works well. However, inference speed is a bottleneck when processing large amounts of data. Since I have four GPUs available, I would like to parallelize the inference process by using all four GPUs, in a way similar to data parallelism.
My idea is to load a full copy of the model on each GPU and have each GPU process 1/4 of the total input data independently. This should allow me to speed up inference by utilizing all available GPUs.
I haven’t found a clear way to achieve this in Ollama. Is there an existing feature or a recommended approach to enable this kind of multi-GPU inference? If not, are there any workarounds or best practices that you would suggest?
Any guidance or insights would be greatly appreciated!
Thanks in advance.

Originally created by @lixiangge on GitHub (Feb 8, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/8947 I’m currently using Ollama with docker to deploy Qwen2.5 on a single GPU, and it works well. However, inference speed is a bottleneck when processing large amounts of data. Since I have four GPUs available, I would like to parallelize the inference process by using all four GPUs, in a way similar to data parallelism. My idea is to load a full copy of the model on each GPU and have each GPU process 1/4 of the total input data independently. This should allow me to speed up inference by utilizing all available GPUs. I haven’t found a clear way to achieve this in Ollama. Is there an existing feature or a recommended approach to enable this kind of multi-GPU inference? If not, are there any workarounds or best practices that you would suggest? Any guidance or insights would be greatly appreciated! Thanks in advance.
GiteaMirror added the feature request label 2026-04-12 17:08:48 -05:00
GiteaMirror commented 2026-04-12 17:08:49 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Feb 8, 2025):

Run four ollama servers, and set CUDA_VISIBLE_DEVICES to a different GPU for each server. Run a proxy server in front to present a unified interface (eg litellm, ollama_proxy, nginx).

<!-- gh-comment-id:2644743656 --> @rick-github commented on GitHub (Feb 8, 2025): Run four ollama servers, and set [`CUDA_VISIBLE_DEVICES`](https://github.com/ollama/ollama/blob/main/docs/gpu.md#gpu-selection) to a different GPU for each server. Run a proxy server in front to present a unified interface (eg [litellm](https://github.com/BerriAI/litellm), [ollama_proxy](https://github.com/ParisNeo/ollama_proxy_server), [nginx](https://github.com/ollama/ollama/issues/8186#issuecomment-2560443545)).
GiteaMirror commented 2026-04-12 17:08:49 -05:00
Author
Owner
Copy Link

@lixiangge commented on GitHub (Feb 16, 2025):

@rick-github
Thanks，it works！

<!-- gh-comment-id:2661188445 --> @lixiangge commented on GitHub (Feb 16, 2025): @rick-github Thanks，it works！
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#5804
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?