mlx-vlm icon indicating copy to clipboard operation
mlx-vlm copied to clipboard

Add support for dots.ocr

Open Garry-TI opened this issue 6 months ago • 3 comments

I wish to request support for dots.ocr model type. I started to convert the model to MLX and made it through the text layers easily as they are Qwen2 type. Then hit the vision tower layers and got stuck when trying to use mlx_lm. I didn't realize there was an mlx_vlm project and though mlx_lm would support it. would love to complete the transition of the dots.ocr model to run native on Apple silicon using MLX.

Garry-TI avatar Aug 25 '25 19:08 Garry-TI

+1 an mlx layout parser would be huge!

dwohlfahrt avatar Aug 27 '25 15:08 dwohlfahrt

Hey @Garry-TI

Awesome work!

Could you open a PR with the progress you made and we will help you finish it.

Or you could wait a bit longer and will do.

Blaizzy avatar Sep 01 '25 17:09 Blaizzy

Hi @Blaizzy great to hear from you. Yes I will add a PR as soon as I resolve this error I created working with the vision tower. :-)

Garry-TI avatar Sep 18 '25 14:09 Garry-TI