github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-07-15 12:48:05 -05:00
Code Issues 16.9k Packages Projects Releases 122 Wiki Activity

[GH-ISSUE #7871] pydantic issue with converted PNG images #5036

New Issue
Closed
opened 2026-04-12 16:07:32 -05:00 by GiteaMirror · 3 comments
GiteaMirror commented 2026-04-12 16:07:32 -05:00
Owner
Copy Link

Originally created by @ibagur on GitHub (Nov 28, 2024).
Original GitHub issue: https://github.com/ollama/ollama/issues/7871

What is the issue?

Directly feeding a PNG image does not work (failed to decode image: image: unknown format), so until recently, I was using the code bellow in order to encode the image file and it used to work fine:

import base64
import io
from PIL import Image
import ollama

def encode_image_to_base64(image_path: str, format: str = "PNG") -> str:
    """Encodes an image file to a base64 string."""
    with Image.open(image_path) as img:
        buffered = io.BytesIO()
        img.save(buffered, format=format)
        return base64.b64encode(buffered.getvalue()).decode('utf-8')

response = ollama.chat(
    model='llama3.2-vision:latest',
    messages=[{
        'role': 'user',
        'content': 'What is in this image?',
        'images': [encode_image_to_base64('image1.png')]
    }]
)

# Print the response
print(response['message']['content'])

But now I get a pydantic serialization error:

PydanticSerializationError: Error calling function `serialize_model`: OSError: [Errno 63] File name too long: 'iVBORw0KGgoAAAANSUhEUgAAAOYAAAGQCAIAAAA1BIuEAAEAAElEQ...

The only workaround I have found to make it work is to directly convert the PNG image into JPG and then feed the JPG image directly to ollama llama3.2-vision:latest model:

def convert_to_jpg(image_path: str) -> bytes:
    """
    Convert image to JPG in memory and return the bytes
    """
    with Image.open(image_path) as img:
        # Convert to RGB if needed
        if img.mode != 'RGB':
            img = img.convert('RGB')
        
        # Save as JPG to memory buffer
        buffered = io.BytesIO()
        img.save(buffered, format='JPEG', quality=85)
        return buffered.getvalue()

response = ollama.chat(
    model='llama3.2-vision:latest',
    messages=[{
        'role': 'user',
        'content': 'What is in this image?',
        'images': [convert_to_jpg('image1.png')]
    }]
)

# Print the response
print(response['message']['content'])

What could be the reason? Any recent update on either ollama, pydantic or pillow libraries? These are the versions I am using in my venv:

ollama            0.4.1
pillow            11.0.0
pydantic          2.10.2
pydantic_core     2.27.1

Thanks for your suggestions!

OS

macOS

GPU

Apple

CPU

Apple

Ollama version

0.4.1

UPDATE:

The particular image I was using was wrong format. Apparently it was a 'webp' image wrongly misnamed as PNG. Now feeding directly a proper PNG image works fine, no need for conversion

Originally created by @ibagur on GitHub (Nov 28, 2024). Original GitHub issue: https://github.com/ollama/ollama/issues/7871 ### What is the issue? Directly feeding a PNG image does not work (`failed to decode image: image: unknown format`), so until recently, I was using the code bellow in order to encode the image file and it used to work fine: ``` import base64 import io from PIL import Image import ollama def encode_image_to_base64(image_path: str, format: str = "PNG") -> str: """Encodes an image file to a base64 string.""" with Image.open(image_path) as img: buffered = io.BytesIO() img.save(buffered, format=format) return base64.b64encode(buffered.getvalue()).decode('utf-8') response = ollama.chat( model='llama3.2-vision:latest', messages=[{ 'role': 'user', 'content': 'What is in this image?', 'images': [encode_image_to_base64('image1.png')] }] ) # Print the response print(response['message']['content']) ``` But now I get a pydantic serialization error: ``` PydanticSerializationError: Error calling function `serialize_model`: OSError: [Errno 63] File name too long: 'iVBORw0KGgoAAAANSUhEUgAAAOYAAAGQCAIAAAA1BIuEAAEAAElEQ... ``` The only workaround I have found to make it work is to directly convert the PNG image into JPG and then feed the JPG image directly to ollama `llama3.2-vision:latest` model: ``` def convert_to_jpg(image_path: str) -> bytes: """ Convert image to JPG in memory and return the bytes """ with Image.open(image_path) as img: # Convert to RGB if needed if img.mode != 'RGB': img = img.convert('RGB') # Save as JPG to memory buffer buffered = io.BytesIO() img.save(buffered, format='JPEG', quality=85) return buffered.getvalue() response = ollama.chat( model='llama3.2-vision:latest', messages=[{ 'role': 'user', 'content': 'What is in this image?', 'images': [convert_to_jpg('image1.png')] }] ) # Print the response print(response['message']['content']) ``` What could be the reason? Any recent update on either `ollama`, `pydantic` or `pillow` libraries? These are the versions I am using in my `venv`: ``` ollama 0.4.1 pillow 11.0.0 pydantic 2.10.2 pydantic_core 2.27.1 ``` Thanks for your suggestions! ### OS macOS ### GPU Apple ### CPU Apple ### Ollama version 0.4.1 UPDATE: The particular image I was using was wrong format. Apparently it was a 'webp' image wrongly misnamed as PNG. Now feeding directly a proper PNG image works fine, no need for conversion
GiteaMirror added the bug label 2026-04-12 16:07:32 -05:00
GiteaMirror commented 2026-04-12 16:07:34 -05:00
Author
Owner
Copy Link

@ibagur commented on GitHub (Nov 28, 2024):

UPDATE:

The particular image I was using was wrong format. Apparently it was a 'webp' image wrongly misnamed as PNG. Now feeding directly a proper PNG image works fine, no need for conversion

<!-- gh-comment-id:2505977090 --> @ibagur commented on GitHub (Nov 28, 2024): UPDATE: The particular image I was using was wrong format. Apparently it was a 'webp' image wrongly misnamed as PNG. Now feeding directly a proper PNG image works fine, no need for conversion
GiteaMirror commented 2026-04-12 16:07:34 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Nov 28, 2024):

Although it works for you, it does highlight a problem. The 'images' field can take either a filename or an encoded image, so in your example you could just use 'images': ['image1.png'] and it would work just fine. What you error revealed is that in order to find out what type of input the chat function has received, it does a stat call on the value of the images field to determine whether it's a file or an encoded image. The stat failed because the string of encoded bytes is too long to be a filename, which results in an exception from serialize_model.

<!-- gh-comment-id:2505987895 --> @rick-github commented on GitHub (Nov 28, 2024): Although it works for you, it does highlight a problem. The 'images' field can take either a filename or an encoded image, so in your example you could just use `'images': ['image1.png']` and it would work just fine. What you error revealed is that in order to find out what type of input the `chat` function has received, it does a `stat` call on the value of the `images` field to determine whether it's a file or an encoded image. The `stat` failed because the string of encoded bytes is too long to be a filename, which results in an exception from `serialize_model`.
GiteaMirror commented 2026-04-12 16:07:35 -05:00
Author
Owner
Copy Link

@ibagur commented on GitHub (Nov 28, 2024):

Although it works for you, it does highlight a problem. The 'images' field can take either a filename or an encoded image,

Yes, in some previous notebooks I had it was taking a base64 encoded image with no issues. That approach is giving error now. I am changing my code to directly feed the filename as you say. Thanks!

<!-- gh-comment-id:2506479669 --> @ibagur commented on GitHub (Nov 28, 2024): > Although it works for you, it does highlight a problem. The 'images' field can take either a filename or an encoded image, Yes, in some previous notebooks I had it was taking a base64 encoded image with no issues. That approach is giving error now. I am changing my code to directly feed the filename as you say. Thanks!
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#5036
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?