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 #1553] customise number of experts in mixtral #47359

New Issue
Open
opened 2026-04-28 03:37:08 -05:00 by GiteaMirror · 4 comments
GiteaMirror commented 2026-04-28 03:37:08 -05:00
Owner
Copy Link

Originally created by @scienlabs on GitHub (Dec 15, 2023).
Original GitHub issue: https://github.com/ollama/ollama/issues/1553

Could you someone provide guidance or documentation on how to adjust the number of experts in mixtral? I'm particularly interested in understanding if there's a way to dynamically adjust this number based on the requirements of different tasks or scenarios.

Originally created by @scienlabs on GitHub (Dec 15, 2023). Original GitHub issue: https://github.com/ollama/ollama/issues/1553 Could you someone provide guidance or documentation on how to adjust the number of experts in mixtral? I'm particularly interested in understanding if there's a way to dynamically adjust this number based on the requirements of different tasks or scenarios.
GiteaMirror commented 2026-04-28 03:37:09 -05:00
Author
Owner
Copy Link

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

I'm not sure what Ollama uses, but for the llama.cpp backend you can override a Key in the model with:

--override-kv KEY=TYPE:VALUE
                        advanced option to override model metadata by key. may be specified multiple times.
                        types: int, float, bool. example: --override-kv tokenizer.ggml.add_bos_token=bool:false

For example I override them using:

--override-kv llama.expert_used_count=int:3

But I think this is not yet supported by MODELFILE.

<!-- gh-comment-id:1863339486 --> @RafaAguilar commented on GitHub (Dec 19, 2023): I'm not sure what Ollama uses, but for the `llama.cpp` [backend](https://github.com/ggerganov/llama.cpp/pull/4406) you can override a Key in the model with: ``` --override-kv KEY=TYPE:VALUE advanced option to override model metadata by key. may be specified multiple times. types: int, float, bool. example: --override-kv tokenizer.ggml.add_bos_token=bool:false ``` For example I override them using: ``` --override-kv llama.expert_used_count=int:3 ``` But I think this is not yet supported by MODELFILE.
GiteaMirror commented 2026-04-28 03:37:10 -05:00
Author
Owner
Copy Link

@scienlabs commented on GitHub (Dec 27, 2023):

how can i do it with ollama? wondering if anyone can help

<!-- gh-comment-id:1870486900 --> @scienlabs commented on GitHub (Dec 27, 2023): how can i do it with ollama? wondering if anyone can help
GiteaMirror commented 2026-04-28 03:37:10 -05:00
Author
Owner
Copy Link

@PLK2 commented on GitHub (May 8, 2024):

Figured it out yet?

<!-- gh-comment-id:2100894627 --> @PLK2 commented on GitHub (May 8, 2024): Figured it out yet?
GiteaMirror commented 2026-04-28 03:37:11 -05:00
Author
Owner
Copy Link

@ColumbusAI commented on GitHub (Aug 2, 2024):

Any update on this?

<!-- gh-comment-id:2264473387 --> @ColumbusAI commented on GitHub (Aug 2, 2024): Any update on this?
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#47359
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?