github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-05 23:53:43 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #1740] PowerInfer Enhancement #994

New Issue
Closed
opened 2026-04-12 10:42:30 -05:00 by GiteaMirror · 1 comment
GiteaMirror commented 2026-04-12 10:42:30 -05:00
Owner
Copy Link

Originally created by @iplayfast on GitHub (Dec 29, 2023).
Original GitHub issue: https://github.com/ollama/ollama/issues/1740

I keep seeing posts about powerinfer https://github.com/SJTU-IPADS/PowerInfer which (if I understand it) keeps often used terms in gpu memory and seldom used terms in cpu memory. This results in an 11x speed up.

It looks like models need to be updated to use this, so it's a pain. BUT.... 11x speed up.

I wonder if the model could be updated automatically so after download it revises it, and stores it.

Anyways, just for interest sake. 11X speed up!!!

Originally created by @iplayfast on GitHub (Dec 29, 2023). Original GitHub issue: https://github.com/ollama/ollama/issues/1740 I keep seeing posts about powerinfer https://github.com/SJTU-IPADS/PowerInfer which (if I understand it) keeps often used terms in gpu memory and seldom used terms in cpu memory. This results in an 11x speed up. It looks like models need to be updated to use this, so it's a pain. BUT.... 11x speed up. I wonder if the model could be updated automatically so after download it revises it, and stores it. Anyways, just for interest sake. **11X speed up!!!**
GiteaMirror commented 2026-04-12 10:42:31 -05:00
Author
Owner
Copy Link

@iplayfast commented on GitHub (Dec 29, 2023):

I see that llama.cpp has already discussed this, and the consensuses is that it's

  1. cherry picked results, actual results will be 3-4x speedup.
  2. code is not complete and needs to be refined.
  3. They think it's interesting and will integrate some of the ideas once it's a bit more stable.

So I guess it's coming from llama.cpp in the distant future.

<!-- gh-comment-id:1872344178 --> @iplayfast commented on GitHub (Dec 29, 2023): I see that llama.cpp has already discussed this, and the consensuses is that it's 1. cherry picked results, actual results will be 3-4x speedup. 2. code is not complete and needs to be refined. 3. They think it's interesting and will integrate some of the ideas once it's a bit more stable. So I guess it's coming from llama.cpp in the distant future.
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#994
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?