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 #3285] gemma accuracy down from 0.128 to 0.129 #64062

New Issue
Open
opened 2026-05-03 16:01:04 -05:00 by GiteaMirror · 3 comments
GiteaMirror commented 2026-05-03 16:01:04 -05:00
Owner
Copy Link

Originally created by @RamiKassouf on GitHub (Mar 21, 2024).
Original GitHub issue: https://github.com/ollama/ollama/issues/3285

What is the issue?

Prompts that were producing the correct results are now producing different (false) outputs

What did you expect to see?

Corretly formatted yaml with correct inputs based on custgom promt

-> Got a yaml with indentation issues and missing fields with wrong structure even though there are json schemas provided

Note that this has also happened on my collegue's pc as well as on a Remote Machine

Steps to reproduce

Upgrade ollama using linux download command provided in website

ollama run gemma

After I rm the model then pull it becomes better for a bit then it get's worse after a while

Are there any recent changes that introduced the issue?

Version Upgrade to 0.129

OS

Linux

Architecture

No response

Platform

WSL2

Ollama version

0.1.29

GPU

Nvidia

GPU info

Quadro P1000

CPU

Intel

Other software

Intel(R) Core(TM) i7-8850H CPU @ 2.60GHz 2.59 GHz

Originally created by @RamiKassouf on GitHub (Mar 21, 2024). Original GitHub issue: https://github.com/ollama/ollama/issues/3285 ### What is the issue? Prompts that were producing the correct results are now producing different (false) outputs ### What did you expect to see? Corretly formatted yaml with correct inputs based on custgom promt -> Got a yaml with indentation issues and missing fields with wrong structure even though there are json schemas provided Note that this has also happened on my collegue's pc as well as on a Remote Machine ### Steps to reproduce Upgrade ollama using linux download command provided in website ollama run gemma After I rm the model then pull it becomes better for a bit then it get's worse after a while ### Are there any recent changes that introduced the issue? Version Upgrade to 0.129 ### OS Linux ### Architecture _No response_ ### Platform WSL2 ### Ollama version 0.1.29 ### GPU Nvidia ### GPU info Quadro P1000 ### CPU Intel ### Other software Intel(R) Core(TM) i7-8850H CPU @ 2.60GHz 2.59 GHz
GiteaMirror added the questionbug labels 2026-05-03 16:01:19 -05:00
GiteaMirror commented 2026-05-03 16:01:23 -05:00
Author
Owner
Copy Link

@BruceMacD commented on GitHub (Mar 25, 2024):

Hi @RamiKassouf, sorry you're seeing this. I've heard in the past about Gemma getting focused on initial prompts and I'm wondering if that could be the case here. Removing the model and pulling it again shouldn't change the behavior, are you able to get better responses after stopping the model and re-running it too?

<!-- gh-comment-id:2018549161 --> @BruceMacD commented on GitHub (Mar 25, 2024): Hi @RamiKassouf, sorry you're seeing this. I've heard in the past about Gemma getting focused on initial prompts and I'm wondering if that could be the case here. Removing the model and pulling it again shouldn't change the behavior, are you able to get better responses after stopping the model and re-running it too?
GiteaMirror commented 2026-05-03 16:01:26 -05:00
Author
Owner
Copy Link

@RamiKassouf commented on GitHub (Mar 26, 2024):

Hi @RamiKassouf, sorry you're seeing this. I've heard in the past about Gemma getting focused on initial prompts and I'm wondering if that could be the case here. Removing the model and pulling it again shouldn't change the behavior, are you able to get better responses after stopping the model and re-running it too?

Yess I do for a little while then it becomes worse
It feels like there's a memory leak or something like that?

<!-- gh-comment-id:2019549914 --> @RamiKassouf commented on GitHub (Mar 26, 2024): > Hi @RamiKassouf, sorry you're seeing this. I've heard in the past about Gemma getting focused on initial prompts and I'm wondering if that could be the case here. Removing the model and pulling it again shouldn't change the behavior, are you able to get better responses after stopping the model and re-running it too? Yess I do for a little while then it becomes worse It feels like there's a memory leak or something like that?
GiteaMirror commented 2026-05-03 16:01:29 -05:00
Author
Owner
Copy Link

@oldgithubman commented on GitHub (Mar 30, 2024):

After updating, ollama is definitely using more memory for me. Even before loading any models

<!-- gh-comment-id:2027908107 --> @oldgithubman commented on GitHub (Mar 30, 2024): After updating, ollama is definitely using more memory for me. Even before loading any models
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 bug question
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#64062
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?