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 #9418] Granite Vision model granite3.2-vision:2b-q8_0 #6142

New Issue
Closed
opened 2026-04-12 17:29:35 -05:00 by GiteaMirror · 2 comments
GiteaMirror commented 2026-04-12 17:29:35 -05:00
Owner
Copy Link

Originally created by @paulstoicasebastian on GitHub (Feb 28, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/9418

What is the issue?

I try OCR on local images from Ollama console (Windows). The response for the second image (and so on) contains the text from the first processed image.

The model granite3.2-vision:latest seams to work fine.

Example (3 different images)

extract text from G:\test4.jpg
Added image 'G:\test4.jpg'

POWER SUPPLY
THE CALCULATOR IS POWERED BY
SOLAR CELL AND A BATTERY FOR
BACKUP.THE DISPLAY WILL DIM WHEN
THE BATTERY NEEDS TO BE REPLACED.
THE " SIGN OF EACH BATTERY MUST

extract text from G:\text2.png
Added image 'G:\text2.png'

POWER SUPPLY
SOLAR CELL AND A BATTERY FOR
BACKUP.THE DISPLAY WILL DIM WHEN
THE BATTERY NEEDS TO BE REPLACED.
THE " SIGN OF EACH BATTERY MUST
SHOW UPWARD WHEN INSERTED.
This is the first line of
this text example.
This is the second line
of the same text.
This is the first line of
this text example.
This is the second line
of the same text.

extract text from G:\text1.png
Added image 'G:\text1.png'

POWER SUPPLY
SOLAR CELL AND A BATTERY MUST
SHOW UPWARD WHEN INSERTED.
This is the first line of
this text example.
This is the second line
of the same text.
It was the best of
times, it was the worst
of wisdom, it was the
age of foolishness...

Relevant log output

OS

Windows

GPU

Nvidia

CPU

Intel

Ollama version

0.5.13-rc1

Originally created by @paulstoicasebastian on GitHub (Feb 28, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/9418 ### What is the issue? I try OCR on local images from Ollama console (Windows). The response for the second image (and so on) contains the text from the first processed image. The model granite3.2-vision:latest seams to work fine. Example (3 different images) >>> extract text from G:\test4.jpg Added image 'G:\test4.jpg' POWER SUPPLY THE CALCULATOR IS POWERED BY SOLAR CELL AND A BATTERY FOR BACKUP.THE DISPLAY WILL DIM WHEN THE BATTERY NEEDS TO BE REPLACED. THE " SIGN OF EACH BATTERY MUST >>> extract text from G:\text2.png Added image 'G:\text2.png' <doc> POWER SUPPLY SOLAR CELL AND A BATTERY FOR BACKUP.THE DISPLAY WILL DIM WHEN THE BATTERY NEEDS TO BE REPLACED. THE " SIGN OF EACH BATTERY MUST SHOW UPWARD WHEN INSERTED. This is the first line of this text example. This is the second line of the same text. This is the first line of this text example. This is the second line of the same text. </doc> >>> extract text from G:\text1.png Added image 'G:\text1.png' <doc> POWER SUPPLY SOLAR CELL AND A BATTERY MUST SHOW UPWARD WHEN INSERTED. This is the first line of this text example. This is the second line of the same text. It was the best of times, it was the worst of wisdom, it was the age of foolishness... </doc> ### Relevant log output ```shell ``` ### OS Windows ### GPU Nvidia ### CPU Intel ### Ollama version 0.5.13-rc1
GiteaMirror added the bug label 2026-04-12 17:29:35 -05:00
GiteaMirror commented 2026-04-12 17:29:36 -05:00
Author
Owner
Copy Link

@jmorganca commented on GitHub (Feb 28, 2025):

Hi @paulstoicasebastian this may be because follow-on images also contain the chat history from the previous ones. Would it be try by creating a new chat for each image? In the meantime, we'll work on improving this so follow-on images don't impact future ones. This may be somewhat model-dependent, but I totally agree having context from one image leak into the next one isn't deseriable.

<!-- gh-comment-id:2691725179 --> @jmorganca commented on GitHub (Feb 28, 2025): Hi @paulstoicasebastian this may be because follow-on images also contain the chat history from the previous ones. Would it be try by creating a new chat for each image? In the meantime, we'll work on improving this so follow-on images don't impact future ones. This may be somewhat model-dependent, but I totally agree having context from one image leak into the next one isn't deseriable.
GiteaMirror commented 2026-04-12 17:29:36 -05:00
Author
Owner
Copy Link

@paulstoicasebastian commented on GitHub (Mar 1, 2025):

Yes, new chat for every pic do resolve this problem.

<!-- gh-comment-id:2692176063 --> @paulstoicasebastian commented on GitHub (Mar 1, 2025): Yes, new chat for every pic do resolve this problem.
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
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#6142
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?