[GH-ISSUE #9574] Llama 3.2 11B Vision Model Fails to Process Images when Projector not Included #68300

Closed
opened 2026-05-04 13:10:42 -05:00 by GiteaMirror · 2 comments
Owner

Originally created by @iifeve on GitHub (Mar 7, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/9574

Vision Model Fails to Process Images in Offline Deployment

thanks for your excellent work!

I'm deploying the model offline by manually downloading and configuring the "llama 3.2 11b vision" model along with its corresponding Modelfile. However, I've encountered an issue where the model cannot properly read/process images. While I suspect there might be an error in the Modelfile configuration, I haven't been able to identify the correct Modelfile specifications for enabling image processing capabilities. Could you please help analyze this issue?

model can not process image:

Image

I also tried with curl:

Image

modelfile I wrote:
Image

Relevant log output


OS

linux

GPU

8*4090

CPU

No response

Ollama version

0.5.13

Originally created by @iifeve on GitHub (Mar 7, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/9574 ### Vision Model Fails to Process Images in Offline Deployment thanks for your excellent work! I'm deploying the model offline by manually downloading and configuring the "llama 3.2 11b vision" model along with its corresponding Modelfile. However, I've encountered an issue where the model cannot properly read/process images. While I suspect there might be an error in the Modelfile configuration, I haven't been able to identify the correct Modelfile specifications for enabling image processing capabilities. Could you please help analyze this issue? model can not process image: ![Image](https://github.com/user-attachments/assets/255bf9e2-5703-4d07-b819-f74d25726d5d) I also tried with curl: ![Image](https://github.com/user-attachments/assets/9b58c484-7085-4713-896a-b93c8c717ea1) modelfile I wrote: ![Image](https://github.com/user-attachments/assets/591579b6-c0fe-4f16-99d3-cc76e54c39c4) ### Relevant log output ```shell ``` ### OS linux ### GPU 8*4090 ### CPU _No response_ ### Ollama version 0.5.13
GiteaMirror added the bug label 2026-05-04 13:10:42 -05:00
Author
Owner

@rick-github commented on GitHub (Mar 7, 2025):

The modelfile doesn't include the projector, the bit that does the actual image processing. See the FROM lines in the output of ollama show --modelfile llama3.2-vision.

<!-- gh-comment-id:2706238263 --> @rick-github commented on GitHub (Mar 7, 2025): The modelfile doesn't include the projector, the bit that does the actual image processing. See the `FROM` lines in the output of `ollama show --modelfile llama3.2-vision`.
Author
Owner

@iifeve commented on GitHub (Mar 18, 2025):

Image
after adding line2, model could processing image now, thanks a lot

<!-- gh-comment-id:2731446143 --> @iifeve commented on GitHub (Mar 18, 2025): ![Image](https://github.com/user-attachments/assets/f924494c-15de-46f9-8fbb-785b8e085380) after adding line2, model could processing image now, thanks a lot
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#68300