Add support for PaddleOCR-VL

[lhr0909]

It would be great to have support for PaddleOCR-VL (especially the latest 1.5 version). I checked there is tutorial from PaddleOCR to run the model via mlx-vlm Python package. I am curious if this is enough to find cues to port into MLXVLM in Swift.

https://www.paddleocr.ai/latest/en/version3.x/pipeline_usage/PaddleOCR-VL-Apple-Silicon.html

Happy to give it a go, but a few pointers will be appreciated. If I get it working I will make sure to post a PR.

[davidkoski]

Check out these docs: https://swiftpackageindex.com/ml-explore/mlx-swift-lm/main/documentation/mlxlmcommon/porting

and you can look at how the existing vlm model ports look compared to the originals. There may be some drift if the python version has been changed after the port and the changes were not reflected here.

Also maybe try out the new skills which includes these porting instructions:

• https://github.com/ml-explore/mlx-swift-lm/tree/main/skills