github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-07-15 20:54:54 -05:00
Code Issues 16.9k Packages Projects Releases 122 Wiki Activity

[GH-ISSUE #7587] Erratic behavior, either can't access to image or cannot recognize what is in the imager #4837

New Issue
Closed
opened 2026-04-12 15:50:06 -05:00 by GiteaMirror · 4 comments
GiteaMirror commented 2026-04-12 15:50:06 -05:00
Owner
Copy Link

Originally created by @MirzaAnoush on GitHub (Nov 9, 2024).
Original GitHub issue: https://github.com/ollama/ollama/issues/7587

What is the issue?

Here the image and the responses. It looks like there is total inconsistency in the answers. I wonder if this is working fine with MacOS? I have tested with ollama-python and got same kind of results.
What is bizarre it is telling that it is text LLM and does not have vision ability!!!
Any thoughts?
I am located in France
Capture d’écran 2024-11-09 à 11 13 15
Capture d’écran 2024-11-09 à 11 11 16

OS

macOS

GPU

No response

CPU

Apple

Ollama version

0.4

Originally created by @MirzaAnoush on GitHub (Nov 9, 2024). Original GitHub issue: https://github.com/ollama/ollama/issues/7587 ### What is the issue? Here the image and the responses. It looks like there is total inconsistency in the answers. I wonder if this is working fine with MacOS? I have tested with ollama-python and got same kind of results. What is bizarre it is telling that it is text LLM and does not have vision ability!!! Any thoughts? I am located in France <img width="1207" alt="Capture d’écran 2024-11-09 à 11 13 15" src="https://github.com/user-attachments/assets/79424fe8-868f-44de-94ea-6a569540b6d5"> <img width="1380" alt="Capture d’écran 2024-11-09 à 11 11 16" src="https://github.com/user-attachments/assets/f4531ea4-cf0b-48d8-92ec-37bbb221b4f5"> ### OS macOS ### GPU _No response_ ### CPU Apple ### Ollama version 0.4
GiteaMirror added the bug label 2026-04-12 15:50:06 -05:00
GiteaMirror commented 2026-04-12 15:50:07 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Nov 9, 2024):

As it said, the model can't access the internet. Download the image.

$ wget https://cdn.paris.fr/paris/2022/09/12/huge-9e714198e72f45cc2145b39c8b0c2f6a.jpg
$ ollama run llama3.2-vision
>>> describe this image: ./huge-9e714198e72f45cc2145b39c8b0c2f6a.jpg
Added image './huge-9e714198e72f45cc2145b39c8b0c2f6a.jpg'
The image showcases a vibrant urban scene, featuring two large murals on 
adjacent buildings that dominate the foreground. The mural on the left is 
predominantly white with black accents and depicts a close-up of a 
person's face, while the one on the right boasts a striking blue hue with 
white details, showcasing an open book accompanied by flowers.

In the background, several other buildings are visible, including a tan 
stone structure with arched windows and a tall tower to the left. A small 
section of blue sky is also present in the upper-right corner. The murals 
appear to be located in a city square or plaza, where people can be seen 
strolling about, adding to the lively atmosphere.

The overall ambiance suggests that this image was captured on a sunny day, 
with a sense of urban vibrancy and cultural richness emanating from the 
scene.
<!-- gh-comment-id:2466189774 --> @rick-github commented on GitHub (Nov 9, 2024): As it said, the model can't access the internet. Download the image. ```console $ wget https://cdn.paris.fr/paris/2022/09/12/huge-9e714198e72f45cc2145b39c8b0c2f6a.jpg $ ollama run llama3.2-vision >>> describe this image: ./huge-9e714198e72f45cc2145b39c8b0c2f6a.jpg Added image './huge-9e714198e72f45cc2145b39c8b0c2f6a.jpg' The image showcases a vibrant urban scene, featuring two large murals on adjacent buildings that dominate the foreground. The mural on the left is predominantly white with black accents and depicts a close-up of a person's face, while the one on the right boasts a striking blue hue with white details, showcasing an open book accompanied by flowers. In the background, several other buildings are visible, including a tan stone structure with arched windows and a tall tower to the left. A small section of blue sky is also present in the upper-right corner. The murals appear to be located in a city square or plaza, where people can be seen strolling about, adding to the lively atmosphere. The overall ambiance suggests that this image was captured on a sunny day, with a sense of urban vibrancy and cultural richness emanating from the scene. ```
GiteaMirror commented 2026-04-12 15:50:07 -05:00
Author
Owner
Copy Link

@MirzaAnoush commented on GitHub (Nov 13, 2024):

Hi,

Thank you so much for your reply.

The reason why I was not able to run the model with pictures on my Mac is because ollama restrict the images to be placed in a folder where ollama has been downloaded. Is this true?

However, this restriction is not when running ollama-python as you can place the image in any folder and it works perfectly as far as you give correct path file to the ollama.

However, the returned information is different than the one executed into the terminal as it adds different type of info and lots of timing data. My question is what is the measurement unit for the timing measurement (there is lots of zeros)

Thank you!

Aram HEKIMIAN
CHAIRMAN & CEO

@.*** @.***>
+33 (0) 4 92 28 88 88
KAPSYS - 694, Avenue Maurice Donat, Parc Haute Technologie, Lot 8
06250 Mougins, FRANCE
www.kapsys.com http://www.kapsys.com/

  https://twitter.com/KapsysGroupTwitter https://twitter.com/KapsysGroup https://www.facebook.com/KAPSYS.fr/  https://www.facebook.com/KAPSYS.fr/Facebook https://www.facebook.com/KAPSYS.fr/  Youtube https://www.youtube.com/channel/UCWmMB2NqLkd25Wm7EIOzh0Q

Le 9 nov. 2024 à 13:05, frob @.***> a écrit :

As it said, the model can't access the internet. Download the image.

$ wget https://cdn.paris.fr/paris/2022/09/12/huge-9e714198e72f45cc2145b39c8b0c2f6a.jpg
$ ollama run llama3.2-vision

describe this image: ./huge-9e714198e72f45cc2145b39c8b0c2f6a.jpg
Added image './huge-9e714198e72f45cc2145b39c8b0c2f6a.jpg'
The image showcases a vibrant urban scene, featuring two large murals on
adjacent buildings that dominate the foreground. The mural on the left is
predominantly white with black accents and depicts a close-up of a
person's face, while the one on the right boasts a striking blue hue with
white details, showcasing an open book accompanied by flowers.

In the background, several other buildings are visible, including a tan
stone structure with arched windows and a tall tower to the left. A small
section of blue sky is also present in the upper-right corner. The murals
appear to be located in a city square or plaza, where people can be seen
strolling about, adding to the lively atmosphere.

The overall ambiance suggests that this image was captured on a sunny day,
with a sense of urban vibrancy and cultural richness emanating from the
scene.
—
Reply to this email directly, view it on GitHub https://github.com/ollama/ollama/issues/7587#issuecomment-2466189774, or unsubscribe https://github.com/notifications/unsubscribe-auth/AKBQ7YPSPYYTMKQAOSBDPCDZ7X263AVCNFSM6AAAAABRO7UGCCVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDINRWGE4DSNZXGQ.
You are receiving this because you authored the thread.

<!-- gh-comment-id:2473683605 --> @MirzaAnoush commented on GitHub (Nov 13, 2024): Hi, Thank you so much for your reply. The reason why I was not able to run the model with pictures on my Mac is because ollama restrict the images to be placed in a folder where ollama has been downloaded. Is this true? However, this restriction is not when running ollama-python as you can place the image in any folder and it works perfectly as far as you give correct path file to the ollama. However, the returned information is different than the one executed into the terminal as it adds different type of info and lots of timing data. My question is what is the measurement unit for the timing measurement (there is lots of zeros) Thank you! Aram HEKIMIAN CHAIRMAN & CEO ***@***.*** ***@***.***> +33 (0) 4 92 28 88 88 KAPSYS - 694, Avenue Maurice Donat, Parc Haute Technologie, Lot 8 06250 Mougins, FRANCE www.kapsys.com <http://www.kapsys.com/>   <https://twitter.com/KapsysGroup>Twitter <https://twitter.com/KapsysGroup> <https://www.facebook.com/KAPSYS.fr/>  <https://www.facebook.com/KAPSYS.fr/>Facebook <https://www.facebook.com/KAPSYS.fr/>  Youtube <https://www.youtube.com/channel/UCWmMB2NqLkd25Wm7EIOzh0Q> > Le 9 nov. 2024 à 13:05, frob ***@***.***> a écrit : > > > As it said, the model can't access the internet. Download the image. > > $ wget https://cdn.paris.fr/paris/2022/09/12/huge-9e714198e72f45cc2145b39c8b0c2f6a.jpg > $ ollama run llama3.2-vision > >>> describe this image: ./huge-9e714198e72f45cc2145b39c8b0c2f6a.jpg > Added image './huge-9e714198e72f45cc2145b39c8b0c2f6a.jpg' > The image showcases a vibrant urban scene, featuring two large murals on > adjacent buildings that dominate the foreground. The mural on the left is > predominantly white with black accents and depicts a close-up of a > person's face, while the one on the right boasts a striking blue hue with > white details, showcasing an open book accompanied by flowers. > > In the background, several other buildings are visible, including a tan > stone structure with arched windows and a tall tower to the left. A small > section of blue sky is also present in the upper-right corner. The murals > appear to be located in a city square or plaza, where people can be seen > strolling about, adding to the lively atmosphere. > > The overall ambiance suggests that this image was captured on a sunny day, > with a sense of urban vibrancy and cultural richness emanating from the > scene. > — > Reply to this email directly, view it on GitHub <https://github.com/ollama/ollama/issues/7587#issuecomment-2466189774>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AKBQ7YPSPYYTMKQAOSBDPCDZ7X263AVCNFSM6AAAAABRO7UGCCVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDINRWGE4DSNZXGQ>. > You are receiving this because you authored the thread. >
GiteaMirror commented 2026-04-12 15:50:08 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Nov 13, 2024):

Hi, Thank you so much for your reply. The reason why I was not able to run the model with pictures on my Mac is because ollama restrict the images to be placed in a folder where ollama has been downloaded. Is this true?

I don't have access to a Mac to verify, but I would be very surprised if this were true. MacOS is unix-like under the GUI, so I would expect normal POSIX behaviour when supplying path names to the ollama client. Can you provide an example of it not working?

However, the returned information is different than the one executed into the terminal as it adds different type of info and lots of timing data. My question is what is the measurement unit for the timing measurement (there is lots of zeros)

This sounds like you are seeing the raw JSON response from the ollama API. The timing information is in nanoseconds, divide by 1000000000 to get seconds. There are many tools for handling JSON, my go-to is jq, which you can install on Macos with:

brew install jq

If you don't have brew installed, see brew.sh.

With jq installed, you can then pretty-print the JSON output and convert timing into seconds:

curl -s localhost:11434/api/generate -d '{"model":"llama3.2-vision","prompt":"hello","stream":false}' \
  | jq '.+{"time":(.eval_duration/1000000000)}|del(.context)'
{
  "model": "llama3.2-vision",
  "created_at": "2024-11-13T15:32:15.858675302Z",
  "response": "Hello! How are you today? Is there something I can help you with or would you like to chat?",
  "done": true,
  "done_reason": "stop",
  "total_duration": 2320789371,
  "load_duration": 34193182,
  "prompt_eval_count": 11,
  "prompt_eval_duration": 97000000,
  "eval_count": 23,
  "eval_duration": 2188000000,
  "time": 2.188
}
<!-- gh-comment-id:2473974836 --> @rick-github commented on GitHub (Nov 13, 2024): > Hi, Thank you so much for your reply. The reason why I was not able to run the model with pictures on my Mac is because ollama restrict the images to be placed in a folder where ollama has been downloaded. Is this true? I don't have access to a Mac to verify, but I would be very surprised if this were true. MacOS is unix-like under the GUI, so I would expect normal POSIX behaviour when supplying path names to the ollama client. Can you provide an example of it not working? > However, the returned information is different than the one executed into the terminal as it adds different type of info and lots of timing data. My question is what is the measurement unit for the timing measurement (there is lots of zeros) This sounds like you are seeing the raw JSON response from the ollama API. The timing information is in nanoseconds, divide by 1000000000 to get seconds. There are many tools for handling JSON, my go-to is `jq`, which you can install on Macos with: ``` brew install jq ``` If you don't have `brew` installed, see [brew.sh](https://brew.sh). With `jq` installed, you can then pretty-print the JSON output and convert timing into seconds: ``` curl -s localhost:11434/api/generate -d '{"model":"llama3.2-vision","prompt":"hello","stream":false}' \ | jq '.+{"time":(.eval_duration/1000000000)}|del(.context)' { "model": "llama3.2-vision", "created_at": "2024-11-13T15:32:15.858675302Z", "response": "Hello! How are you today? Is there something I can help you with or would you like to chat?", "done": true, "done_reason": "stop", "total_duration": 2320789371, "load_duration": 34193182, "prompt_eval_count": 11, "prompt_eval_duration": 97000000, "eval_count": 23, "eval_duration": 2188000000, "time": 2.188 } ```
GiteaMirror commented 2026-04-12 15:50:08 -05:00
Author
Owner
Copy Link

@pdevine commented on GitHub (Nov 13, 2024):

As @rick-github mentioned, Ollama can't parse the URL that you provided because it doesn't talk to the internet. The LLM was inferring from the URL what you were asking it and spitting out nonsense (clearly that's next to the Pompidou and not the Arc de Triomphe!). You'll need to specify a local path of MacOS to whatever image that you want to read in. And yes, it can read from folders other than Downloads/, although everything depends on what security settings you've used w/ your machine.

I'll go ahead and close the issue.

<!-- gh-comment-id:2474813015 --> @pdevine commented on GitHub (Nov 13, 2024): As @rick-github mentioned, Ollama can't parse the URL that you provided because it doesn't talk to the internet. The LLM was inferring from the URL what you were asking it and spitting out nonsense (clearly that's next to the Pompidou and not the Arc de Triomphe!). You'll need to specify a local path of MacOS to whatever image that you want to read in. And yes, it can read from folders other than `Downloads/`, although everything depends on what security settings you've used w/ your machine. I'll go ahead and close the issue.
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#4837
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?